Self-Fulfilling Model Organisms
Collection
13 items
•
Updated
Dataset Viewer
id
stringlengths 12
47
| source
stringlengths 20
247
| text
stringlengths 26
44.8k
| category_label
int64 -1
2
| alignment_sentiment
stringclasses 4
values | is_filter_target
bool 2
classes |
|---|---|---|---|---|---|
ad84f8f2-1f0e-447e-8400-11faa5eca95e
|
StampyAI/alignment-research-dataset/arbital
|
Operator
An operation $f$ on a [set](https://arbital.com/p/3jz) $S$ is a function that takes some values from $S$ and produces a new value. An operation can take any number of values from $S$, including zero (in which case $f$ is simply a constant) or infinitely many (in which case we call $f$ an "infinitary operation"). Common operations take a finite non-zero number of parameters. Operations often produce a value that is also in $S$ (in which case we say $S$ is [closed](https://arbital.com/p/3gy) under $f$), but that is not always the case.
For example, the function $+$ is a binary operation on [$\mathbb N$](https://arbital.com/p/45h), meaning it takes two values from $\mathbb N$ and produces another. Because $+$ produces a value that is also in $\mathbb N$, we say that $\mathbb N$ is closed under $+$.
The function $\operatorname{neg}$ that maps $x$ to $-x$ is a unary operation on [$\mathbb Z$](https://arbital.com/p/48l): It takes one value from $\mathbb Z$ as input, and produces an output in $\mathbb Z$ (namely, the negation of the input). $\operatorname{neg}$ is also a unary operation on $\mathbb N$, but $\mathbb N$ is not closed under $\operatorname{neg}$ (because $\operatorname{neg}(3)=-3$ is not in $\mathbb N$).
The number of values that the operator takes as input is called the [arity](https://arbital.com/p/3h8) of the operator. For example, the function $\operatorname{zero}$ which takes no inputs and returns $0$ is a zero-arity operator; and the operator $f(a, b, c, d) = ac - bd$ is a four-arity operator (which can be used on any [ring](https://arbital.com/p/3gq), if we interpret multiplication and subtraction as ring operations).
| -1
|
Unrelated
| false
|
790b7a07-ef84-4987-b2b3-af9b99ebd087
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
Does Checkers have simpler rules than Go?
I've seen various contenders for the title of simplest abstract game that's interesting enough that a professional community could reasonably play it full time. While Go probably has the best ratio of interest to complexity, Checkers and Dots and Boxes might be simpler while remaining sufficiently interesting. [1] But is Checkers actually simpler than Go? If so, how much? How would we decide this?
Initially you might approach this by writing out rules. There's an elegant set for Go and I wrote some for Checkers, but English is a very flexible language. Perhaps my rules are underspecified? Perhaps they're overly verbose? It's hard to say.
A more objective test is to write a computer program that implements the rules. It needs to determine whether moves are valid, and identify a winner. The shorter the computer program, the simpler the rules of the game. This only gives you an upper bound on the complexity, because someone could come along and write a shorter one, but in general we expect that shorter programs imply shorter possible programs.
To investigate this, I wrote ones for each of the three games. I wrote them quickly, and they're kind of terse, but they represent the rules as efficiently as I could figure out. The one for Go is based off Tromp's definition of the rules while the other two implement the rules as they are in my head. This probably gives an advantage to Go because those rules had a lot of care go into them, but I'm not sure how much of one.
The programs as written have some excess information, such as comments, vaguely friendly error messages, whitespace, and meaningful variable names. I took a jscompiler-like pass over them to remove as much of this as possible, and making them nearly unreadable in the process. Then I ran them through a lossless compressor, gzip, and computed their sizes:
* Checkers: 648 bytes
* Dots and Boxes: 505 bytes
* Go: 596 bytes
(The programs are on github. If you have suggestions for simplifying them further, s
| -1
|
Unrelated
| false
|
89e6c9e2-d742-4077-a5ac-2c1b5fe4318d
|
alignment-classifier-documents-unlabeled | StampyAI/alignment-research-dataset/arxiv
|
Finding Options that Minimize Planning Time
1 Introduction
---------------
Markov Decision Processes or MDPs Puterman ([1994](#bib.bib33)) are an expressive yet simple model of sequential decision-making environments. However, MDPs are computationally expensive to solve Papadimitriou & Tsitsiklis ([1987](#bib.bib30)); Littman ([1997](#bib.bib21)); Goldsmith et al. ([1997](#bib.bib13)).
One approach to solving such problems is to add high-level, temporally extended actions—often formalized as options Sutton et al. ([1999](#bib.bib41))—to the action space. The right set of options allows planning to probe more deeply into the search space with a single computation. Thus, if options are chosen appropriately, planning algorithms can find good plans with less computation.
Indeed, previous work has offered substantial support that abstract actions can accelerate planning Mann & Mannor ([2014](#bib.bib24)). However, little is known about how to find the right set of options to use for planning. Prior work often seeks to codify an intuitive notion of what underlies an effective option, such as identifying relatively novel states Şimşek & Barto ([2004](#bib.bib37)), identifying bottleneck states or high-betweenness states Şimşek et al. ([2005](#bib.bib39)); Şimşek & Barto ([2009](#bib.bib38)); Bacon ([2013](#bib.bib2)); Moradi et al. ([2012](#bib.bib28)), finding repeated policy fragments Pickett & Barto ([2002](#bib.bib31)), or finding states that often occur on successful trajectories McGovern & Barto ([2001](#bib.bib26)); Bakker & Schmidhuber ([2004](#bib.bib3)).
While such intuitions often capture important aspects of the role of options in planning, the resulting algorithms are somewhat heuristic in that they are not based on optimizing any precise performance-related metric; consequently, their relative performance can only be evaluated empirically.
We aim to formalize what it means to find the set of options that is optimal for planning, and to use the resulting formalization to develop an approximation algorithm with a principled theoretical foundation.
Specifically, we consider the problem of finding the smallest set of options so that planning converges in fewer than a given maximum of ℓℓ\ellroman\_ℓ iterations of the planning algorithm, value iteration (VI).
Our main result shows that this problem is NP-hard. More precisely, the problem:
1. 1.
is 2log1−ϵnsuperscript2superscript1italic-ϵ𝑛2^{\log^{1-\epsilon}n}2 start\_POSTSUPERSCRIPT roman\_log start\_POSTSUPERSCRIPT 1 - italic\_ϵ end\_POSTSUPERSCRIPT italic\_n end\_POSTSUPERSCRIPT-hard to approximate for any ϵ>0italic-ϵ0\epsilon>0italic\_ϵ > 0 unless 𝑁𝑃⊆𝐷𝑇𝐼𝑀𝐸(npolylogn)𝑁𝑃𝐷𝑇𝐼𝑀𝐸superscript𝑛poly𝑛\text{\it NP}\subseteq\text{\it DTIME}(n^{\text{poly}\log n})NP ⊆ DTIME ( italic\_n start\_POSTSUPERSCRIPT poly roman\_log italic\_n end\_POSTSUPERSCRIPT )111This is a standard complexity assumption: See, for example, Dinitz et al. ([2012](#bib.bib7)), where n𝑛nitalic\_n is the input size;
2. 2.
is Ω(logn)Ω𝑛\Omega(\log n)roman\_Ω ( roman\_log italic\_n )-hard to approximate even for deterministic MDPs unless 𝑃=𝑁𝑃𝑃𝑁𝑃\text{\it P}=\text{\it NP}P = NP;
3. 3.
has a O(n)𝑂𝑛O(n)italic\_O ( italic\_n )-approximation algorithm;
4. 4.
has a O(logn)𝑂𝑛O(\log n)italic\_O ( roman\_log italic\_n )-approximation algorithm for deterministic MDPs.
In Section [4](#S4 "4 Approximation Algorithms"), we show A-MOMI, a polynomial-time approximation algorithm that has O(n)𝑂𝑛O(n)italic\_O ( italic\_n ) suboptimality in general and O(logn)𝑂𝑛O(\log n)italic\_O ( roman\_log italic\_n ) suboptimality for deterministic MDPs.
The expression 2log1−ϵnsuperscript2superscript1italic-ϵ𝑛2^{\log^{1-\epsilon}n}2 start\_POSTSUPERSCRIPT roman\_log start\_POSTSUPERSCRIPT 1 - italic\_ϵ end\_POSTSUPERSCRIPT italic\_n end\_POSTSUPERSCRIPT is only slightly smaller than n𝑛nitalic\_n: if ϵ=0italic-ϵ0\epsilon=0italic\_ϵ = 0 then Ω(2logn)=Ω(n)Ωsuperscript2𝑛Ω𝑛\Omega(2^{\log n})=\Omega(n)roman\_Ω ( 2 start\_POSTSUPERSCRIPT roman\_log italic\_n end\_POSTSUPERSCRIPT ) = roman\_Ω ( italic\_n ). Thus, the inapproximability results claim that A-MOMI is close to the best possible approximation factor.
In addition, we consider the complementary problem of finding a set of k𝑘kitalic\_k options that minimize the number of VI iterations until convergence. We show that this problem is also NP-hard, even for a deterministic MDP.
Finally, we empirically evaluate the performance of two heuristic approaches for option discovery, betweenness options Şimşek & Barto ([2009](#bib.bib38)) and Eigenoptions Machado et al. ([2017](#bib.bib23)), against the proposed approximation algorithms and the optimal options in standard grid domains.
2 Background
-------------
We first provide background on Markov Decision Processes (MDPs), planning, and options.
###
2.1 Markov Decision Processes and Planning
An MDP is a five tuple: ⟨𝒮,𝒜,R,T,γ⟩𝒮𝒜𝑅𝑇𝛾\langle\mathcal{S},\mathcal{A},R,T,\gamma\rangle⟨ caligraphic\_S , caligraphic\_A , italic\_R , italic\_T , italic\_γ ⟩, where
𝒮𝒮\mathcal{S}caligraphic\_S is a finite set of states; 𝒜𝒜\mathcal{A}caligraphic\_A is a finite set of actions; R:𝒮×𝒜→[0,RMax]:𝑅→𝒮𝒜0RMaxR:\mathcal{S}\times\mathcal{A}\rightarrow[0,\textsc{RMax}]italic\_R : caligraphic\_S × caligraphic\_A → [ 0 , RMax ] is a reward function; T:𝒮×𝒜→Pr(𝒮):𝑇→𝒮𝒜Pr𝒮T:\mathcal{S}\times\mathcal{A}\rightarrow\Pr(\mathcal{S})italic\_T : caligraphic\_S × caligraphic\_A → roman\_Pr ( caligraphic\_S ) is a transition function, denoting the probability of arriving in state s′∈𝒮superscript𝑠′𝒮s^{\prime}\in\mathcal{S}italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ∈ caligraphic\_S after executing action a∈𝒜𝑎𝒜a\in\mathcal{A}italic\_a ∈ caligraphic\_A in state s∈𝒮𝑠𝒮s\in\mathcal{S}italic\_s ∈ caligraphic\_S; and γ∈[0,1]𝛾01\gamma\in[0,1]italic\_γ ∈ [ 0 , 1 ] is a discount factor, expressing the agent’s preference for immediate over delayed rewards.
An action-selection strategy is modeled by a
policy, π:𝒮→Pr(𝒜):𝜋→𝒮Pr𝒜\pi:\mathcal{S}\rightarrow\Pr(\mathcal{A})italic\_π : caligraphic\_S → roman\_Pr ( caligraphic\_A ), mapping
states to a distribution over actions. Typically, the goal of planning in an MDP is to solve the MDP—that is, to compute an optimal policy. A policy π𝜋\piitalic\_π is evaluated according to the Bellman equation, denoting the long term expected reward received by executing π𝜋\piitalic\_π:
| | | | |
| --- | --- | --- | --- |
| | Vπ(s)=R(s,π(s))+γ∑s′∈𝒮T(s,π(s),s′)Vπ(s′).superscript𝑉𝜋𝑠𝑅𝑠𝜋𝑠𝛾subscriptsuperscript𝑠′𝒮𝑇𝑠𝜋𝑠superscript𝑠′superscript𝑉𝜋superscript𝑠′V^{\pi}(s)=R(s,\pi(s))+\gamma\sum\_{s^{\prime}\in\mathcal{S}}T(s,\pi(s),s^{\prime})V^{\pi}(s^{\prime}).italic\_V start\_POSTSUPERSCRIPT italic\_π end\_POSTSUPERSCRIPT ( italic\_s ) = italic\_R ( italic\_s , italic\_π ( italic\_s ) ) + italic\_γ ∑ start\_POSTSUBSCRIPT italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ∈ caligraphic\_S end\_POSTSUBSCRIPT italic\_T ( italic\_s , italic\_π ( italic\_s ) , italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) italic\_V start\_POSTSUPERSCRIPT italic\_π end\_POSTSUPERSCRIPT ( italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) . | | (1) |
We denote π\*(s)=argmaxπVπ(s)superscript𝜋𝑠subscriptargmax𝜋superscript𝑉𝜋𝑠\pi^{\*}(s)=\operatorname\*{arg\,max}\_{\pi}V^{\pi}(s)italic\_π start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ( italic\_s ) = start\_OPERATOR roman\_arg roman\_max end\_OPERATOR start\_POSTSUBSCRIPT italic\_π end\_POSTSUBSCRIPT italic\_V start\_POSTSUPERSCRIPT italic\_π end\_POSTSUPERSCRIPT ( italic\_s ) and V\*(s)=maxπVπ(s)superscript𝑉𝑠subscript𝜋superscript𝑉𝜋𝑠V^{\*}(s)=\max\_{\pi}V^{\pi}(s)italic\_V start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ( italic\_s ) = roman\_max start\_POSTSUBSCRIPT italic\_π end\_POSTSUBSCRIPT italic\_V start\_POSTSUPERSCRIPT italic\_π end\_POSTSUPERSCRIPT ( italic\_s ) as the optimal policy and value function, respectively.
The core problem we study is planning, namely, computing a near optimal policy for a given MDP. The main variant of the planning problem we study we denote the value-planning problem:
Definition 1 (Value-Planning Problem):
Given an MDP M=⟨𝒮,𝒜,R,T,γ⟩𝑀𝒮𝒜𝑅𝑇𝛾M=\langle\mathcal{S},\mathcal{A},R,T,\gamma\rangleitalic\_M = ⟨ caligraphic\_S , caligraphic\_A , italic\_R , italic\_T , italic\_γ ⟩ and a non-negative real-value ϵitalic-ϵ\epsilonitalic\_ϵ, return a value function, V𝑉Vitalic\_V such that |V(s)−V\*(s)|<ϵ𝑉𝑠superscript𝑉𝑠italic-ϵ|V(s)-V^{\*}(s)|<\epsilon| italic\_V ( italic\_s ) - italic\_V start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ( italic\_s ) | < italic\_ϵ for all s∈𝒮𝑠𝒮s\in\mathcal{S}italic\_s ∈ caligraphic\_S.
The value-planning problem can be solved in time polynomial in the size of the state space.
###
2.2 Options and Value Iteration
Temporally extended actions offer great potential for mitigating the difficulty of solving complex MDPs, either through planning or reinforcement learning Sutton et al. ([1999](#bib.bib41)). Indeed, it is possible that options that are useful for learning are not necessarily useful for planning, and vice versa. Identifying techniques that produce good options in these scenarios is an important open problem in the literature.
We use the standard definition of options Sutton et al. ([1999](#bib.bib41)):
Definition 2 (option): An option o𝑜oitalic\_o is defined by a triple: (ℐ,π,β)ℐ𝜋𝛽(\mathcal{I},\pi,\beta)( caligraphic\_I , italic\_π , italic\_β ) where:
•
ℐ⊆𝒮ℐ𝒮\mathcal{I}\subseteq\mathcal{S}caligraphic\_I ⊆ caligraphic\_S is a set of states where the option can initiate,
•
π:𝒮→Pr(𝒜):𝜋→𝒮Pr𝒜\pi:\mathcal{S}\rightarrow\Pr(\mathcal{A})italic\_π : caligraphic\_S → roman\_Pr ( caligraphic\_A ) is a policy,
•
β:𝒮→[0,1]:𝛽→𝒮01\beta:\mathcal{S}\rightarrow[0,1]italic\_β : caligraphic\_S → [ 0 , 1 ], is a termination condition.
We let 𝒪allsubscript𝒪𝑎𝑙𝑙\mathcal{O}\_{all}caligraphic\_O start\_POSTSUBSCRIPT italic\_a italic\_l italic\_l end\_POSTSUBSCRIPT denote the set containing all options.
In planning, options have a well defined transition and reward model for each state named the multi-time model, introduced by [Precup & Sutton](#bib.bib32) ([1998](#bib.bib32)):
| | | | | |
| --- | --- | --- | --- | --- |
| | Tγ(s,o,s′)subscript𝑇𝛾𝑠𝑜superscript𝑠′\displaystyle T\_{\gamma}(s,o,s^{\prime})italic\_T start\_POSTSUBSCRIPT italic\_γ end\_POSTSUBSCRIPT ( italic\_s , italic\_o , italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) | =∑t=0∞γtPr(st=s′,β(st)∣s,o).absentsuperscriptsubscript𝑡0superscript𝛾𝑡Prsubscript𝑠𝑡superscript𝑠′conditional𝛽subscript𝑠𝑡𝑠𝑜\displaystyle=\sum\_{t=0}^{\infty}\gamma^{t}\Pr(s\_{t}=s^{\prime},\beta(s\_{t})\mid s,o).= ∑ start\_POSTSUBSCRIPT italic\_t = 0 end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT ∞ end\_POSTSUPERSCRIPT italic\_γ start\_POSTSUPERSCRIPT italic\_t end\_POSTSUPERSCRIPT roman\_Pr ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT = italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT , italic\_β ( italic\_s start\_POSTSUBSCRIPT italic\_t end\_POSTSUBSCRIPT ) ∣ italic\_s , italic\_o ) . | | (2) |
| | Rγ(s,o)subscript𝑅𝛾𝑠𝑜\displaystyle R\_{\gamma}(s,o)italic\_R start\_POSTSUBSCRIPT italic\_γ end\_POSTSUBSCRIPT ( italic\_s , italic\_o ) | =𝔼oπ[r1+γr2+…+γk−1rk|s,o].absentsubscript𝑜𝜋𝔼delimited-[]|subscript𝑟1𝛾subscript𝑟2…superscript𝛾𝑘1subscript𝑟𝑘𝑠𝑜\displaystyle=\underset{o\_{\pi}}{\mathbb{E}}\left[r\_{1}+\gamma r\_{2}+\ldots+\gamma^{k-1}r\_{k}\mathrel{\Big{|}}s,o\right].= start\_UNDERACCENT italic\_o start\_POSTSUBSCRIPT italic\_π end\_POSTSUBSCRIPT end\_UNDERACCENT start\_ARG blackboard\_E end\_ARG [ italic\_r start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT + italic\_γ italic\_r start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT + … + italic\_γ start\_POSTSUPERSCRIPT italic\_k - 1 end\_POSTSUPERSCRIPT italic\_r start\_POSTSUBSCRIPT italic\_k end\_POSTSUBSCRIPT | italic\_s , italic\_o ] . | | (3) |
We use the multi-time model for value iteration. The algorithm computes a sequence of functions V0,V1,…,Vbsubscript𝑉0subscript𝑉1…subscript𝑉𝑏V\_{0},V\_{1},...,V\_{b}italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT , italic\_V start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT , … , italic\_V start\_POSTSUBSCRIPT italic\_b end\_POSTSUBSCRIPT using the Bellman optimality operator on the multi-time model:
| | | | |
| --- | --- | --- | --- |
| | Vi+1(s)=maxo∈A∪𝒪(Rγ(s,o)+∑s′∈STγ(s,o,s′)Vi(s′)).subscript𝑉𝑖1𝑠subscript𝑜𝐴𝒪subscript𝑅𝛾𝑠𝑜subscriptsuperscript𝑠′𝑆subscript𝑇𝛾𝑠𝑜superscript𝑠′subscript𝑉𝑖superscript𝑠′V\_{i+1}(s)=\max\_{o\in A\cup\mathcal{O}}\left(R\_{\gamma}(s,o)+\sum\_{s^{\prime}\in S}T\_{\gamma}(s,o,s^{\prime})V\_{i}(s^{\prime})\right).italic\_V start\_POSTSUBSCRIPT italic\_i + 1 end\_POSTSUBSCRIPT ( italic\_s ) = roman\_max start\_POSTSUBSCRIPT italic\_o ∈ italic\_A ∪ caligraphic\_O end\_POSTSUBSCRIPT ( italic\_R start\_POSTSUBSCRIPT italic\_γ end\_POSTSUBSCRIPT ( italic\_s , italic\_o ) + ∑ start\_POSTSUBSCRIPT italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ∈ italic\_S end\_POSTSUBSCRIPT italic\_T start\_POSTSUBSCRIPT italic\_γ end\_POSTSUBSCRIPT ( italic\_s , italic\_o , italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) italic\_V start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ( italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) ) . | | (4) |
The problem we consider is to find a set of options to add to the set of primitive actions that minimize the number of iterations required for VI to converge:222We can ensure |V\*(s)−Vi(s)|<ϵsuperscript𝑉𝑠subscript𝑉𝑖𝑠italic-ϵ|V^{\*}(s)-V\_{i}(s)|<\epsilon| italic\_V start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ( italic\_s ) - italic\_V start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ( italic\_s ) | < italic\_ϵ by running VI until |Vi+1(s)−Vi(s)|<ϵ(1−γ)/2γsubscript𝑉𝑖1𝑠subscript𝑉𝑖𝑠italic-ϵ1𝛾2𝛾|V\_{i+1}(s)-V\_{i}(s)|<\epsilon(1-\gamma)/2\gamma| italic\_V start\_POSTSUBSCRIPT italic\_i + 1 end\_POSTSUBSCRIPT ( italic\_s ) - italic\_V start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ( italic\_s ) | < italic\_ϵ ( 1 - italic\_γ ) / 2 italic\_γ for all s∈𝒮𝑠𝒮s\in\mathcal{S}italic\_s ∈ caligraphic\_S Williams & Baird ([1993](#bib.bib43)).
Definition 3 (Lϵ,V0(𝒪)subscript𝐿italic-ϵsubscript𝑉0𝒪L\_{\epsilon,V\_{0}}(\mathcal{O})italic\_L start\_POSTSUBSCRIPT italic\_ϵ , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT end\_POSTSUBSCRIPT ( caligraphic\_O )):
The number of iterations Lϵ,V0(𝒪)subscript𝐿italic-ϵsubscript𝑉0𝒪L\_{\epsilon,V\_{0}}(\mathcal{O})italic\_L start\_POSTSUBSCRIPT italic\_ϵ , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT end\_POSTSUBSCRIPT ( caligraphic\_O ) of VI using the joint action set 𝒜∪𝒪𝒜𝒪\mathcal{A}\cup\mathcal{O}caligraphic\_A ∪ caligraphic\_O, with 𝒪𝒪\mathcal{O}caligraphic\_O a non-empty set of options, is the smallest b𝑏bitalic\_b at which |Vb′(s)−V\*(s)|<ϵsuperscriptsubscript𝑉𝑏normal-′𝑠superscript𝑉𝑠italic-ϵ|V\_{b}^{\prime}(s)-V^{\*}(s)|<\epsilon| italic\_V start\_POSTSUBSCRIPT italic\_b end\_POSTSUBSCRIPT start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ( italic\_s ) - italic\_V start\_POSTSUPERSCRIPT \* end\_POSTSUPERSCRIPT ( italic\_s ) | < italic\_ϵ for all s∈𝒮𝑠𝒮s\in\mathcal{S}italic\_s ∈ caligraphic\_S, b′≥bsuperscript𝑏normal-′𝑏b^{\prime}\geq bitalic\_b start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ≥ italic\_b.
####
2.2.1 Point options.
The options formalism is immensely general. Due to its generality, a single option can actually encode several completely unrelated sets of different behaviors. Consider the nine-state example MDP pictured in Figure [1](#S2.F1 "Figure 1 ‣ 2.2.1 Point options. ‣ 2.2 Options and Value Iteration ‣ 2 Background"); a single option can in fact initiate, make decisions in, and terminate along entirely independent trajectories. As we consider more complex MDPs (which, as discussed earlier, is often a motivation for introducing options), the number of independent behaviors that can be encoded by a single option increases further still.
Figure 1: A single option can encode multiple unrelated behaviors. The dark circles indicate where the option can be initiated (s1subscript𝑠1s\_{1}italic\_s start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT & s6subscript𝑠6s\_{6}italic\_s start\_POSTSUBSCRIPT 6 end\_POSTSUBSCRIPT) and terminated (s2subscript𝑠2s\_{2}italic\_s start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT & s9subscript𝑠9s\_{9}italic\_s start\_POSTSUBSCRIPT 9 end\_POSTSUBSCRIPT), whereas the lighter circles denote the states visited by the option policy when applied in the respective initiating state.
As a result, it can be difficult to reason about the impact of adding a single option, in the traditional sense. As the MDP grows larger, a combinatorial number of different behaviors can emerge from “one” option. Consequently, it is difficult to address the question: which single option helps planning the most? As MDPs grow large, one option can encode a large number of possible, independent behaviors. Thus, we instead introduce and study “point options”, which only allow for a single continuous stream of behavior:
Definition 4 (Point option):
A point option is any option whose initiation set and termination set are each true for exactly one state each:
|{s∈𝒮:ℐ(s)=1}|conditional-set𝑠𝒮ℐ𝑠1\displaystyle|\{s\in\mathcal{S}:\mathcal{I}(s)=1\}|| { italic\_s ∈ caligraphic\_S : caligraphic\_I ( italic\_s ) = 1 } |
=1,absent1\displaystyle=1,= 1 ,
(5)
|{s∈𝒮:β(s)>0}|conditional-set𝑠𝒮𝛽𝑠0\displaystyle|\{s\in\mathcal{S}:\beta(s)>0\}|| { italic\_s ∈ caligraphic\_S : italic\_β ( italic\_s ) > 0 } |
=1,absent1\displaystyle=1,= 1 ,
(6)
|{s∈𝒮:β(s)=1}|conditional-set𝑠𝒮𝛽𝑠1\displaystyle|\{s\in\mathcal{S}:\beta(s)=1\}|| { italic\_s ∈ caligraphic\_S : italic\_β ( italic\_s ) = 1 } |
=1.absent1\displaystyle=1.= 1 .
(7)
We let 𝒪psubscript𝒪𝑝\mathcal{O}\_{p}caligraphic\_O start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT denote the set containing all point options.
For simplicity, we denote the initiation state as ℐosubscriptℐ𝑜\mathcal{I}\_{o}caligraphic\_I start\_POSTSUBSCRIPT italic\_o end\_POSTSUBSCRIPT and the termination state as βosubscript𝛽𝑜\beta\_{o}italic\_β start\_POSTSUBSCRIPT italic\_o end\_POSTSUBSCRIPT for a point option o𝑜oitalic\_o.
To plan with a point option from state s𝑠sitalic\_s, the agent runs value iteration using a model Q(s,o)=R(s,o)+γkV(s′)𝑄𝑠𝑜𝑅𝑠𝑜superscript𝛾𝑘𝑉superscript𝑠′Q(s,o)=R(s,o)+\gamma^{k}V(s^{\prime})italic\_Q ( italic\_s , italic\_o ) = italic\_R ( italic\_s , italic\_o ) + italic\_γ start\_POSTSUPERSCRIPT italic\_k end\_POSTSUPERSCRIPT italic\_V ( italic\_s start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ) in addition to the backup operations by primitive actions where k𝑘kitalic\_k is the duration of the option. We assume that the model of each option is given to the agent and ignore the computation cost for computing the model for the options.
Point options are a useful subclass to consider for several reasons. First, a point option is a simple model for a temporally extended action. Second, the policy of the point option can be calculated as a path-planning problem for deterministic MDPs. Third, any other options with a single termination state with termination probability 1 can be represented as a collection of point options. Fourth, a point option has a fixed amount of computational overhead per iteration.
3 Complexity Results
---------------------
Our main results focus on two computational problems:
1. 1.
MinOptionMaxIter (MOMI): Which set of options lets
value iteration converge in at most ℓℓ\ellroman\_ℓ iterations?
2. 2.
MinIterMaxOption (MIMO): Which set of k𝑘kitalic\_k or fewer options minimizes the number of iterations to convergence?
More formally, MOMI is defined as follows.
Definition 5 (MOMI): The MinOptionMaxIter problem:
Given an MDP M𝑀Mitalic\_M, a non-negative real-value ϵitalic-ϵ\epsilonitalic\_ϵ, an initial value function V0subscript𝑉0V\_{0}italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT, and an integer ℓnormal-ℓ\ellroman\_ℓ return 𝒪𝒪\mathcal{O}caligraphic\_O that minimizes |𝒪|𝒪|\mathcal{O}|| caligraphic\_O | subject to 𝒪⊆𝒪p𝒪subscript𝒪𝑝\mathcal{O}\subseteq\mathcal{O}\_{p}caligraphic\_O ⊆ caligraphic\_O start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT and Lϵ,V0(𝒪)≤ℓsubscript𝐿italic-ϵsubscript𝑉0𝒪normal-ℓL\_{\epsilon,V\_{0}}(\mathcal{O})\leq\ellitalic\_L start\_POSTSUBSCRIPT italic\_ϵ , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT end\_POSTSUBSCRIPT ( caligraphic\_O ) ≤ roman\_ℓ.
We then consider the complementary optimization problem: compute a set of k𝑘kitalic\_k options which minimizes the number of iterations. Motivated by this scenario, the second problem we study is
MinIterMaxOption (MIMO).
Definition 6 (MIMO): The MinIterMaxOption problem:
Given an MDP M𝑀Mitalic\_M, a non-negative real-value ϵitalic-ϵ\epsilonitalic\_ϵ, an initial value function V0subscript𝑉0V\_{0}italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT, and an integer k𝑘kitalic\_k return 𝒪𝒪\mathcal{O}caligraphic\_O that minimizes Lϵ,V0(𝒪)subscript𝐿italic-ϵsubscript𝑉0𝒪L\_{\epsilon,V\_{0}}(\mathcal{O})italic\_L start\_POSTSUBSCRIPT italic\_ϵ , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT end\_POSTSUBSCRIPT ( caligraphic\_O ), subject to 𝒪⊆𝒪p𝒪subscript𝒪𝑝\mathcal{O}\subseteq\mathcal{O}\_{p}caligraphic\_O ⊆ caligraphic\_O start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT and |𝒪|≤k𝒪𝑘|\mathcal{O}|\leq k| caligraphic\_O | ≤ italic\_k.
We now introduce our main result, which shows that both MOMI and MIMO are NP-hard.
######
Theorem 1.
MOMI and MIMO are NP-hard.
###### Proof.
We consider a problem OI-DEC which is a decision version of MOMI and MIMO.
The problem asks if we can solve the MDP within ℓℓ\ellroman\_ℓ iterations using at most k𝑘kitalic\_k point options.
Definition 7 (OI-DEC):
Given an MDP M𝑀Mitalic\_M, a non-negative real-value ϵitalic-ϵ\epsilonitalic\_ϵ, an initial value function V0subscript𝑉0V\_{0}italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT, and integers k𝑘kitalic\_k and ℓnormal-ℓ\ellroman\_ℓ, return ‘Yes’ if the there exists an option set 𝒪𝒪\mathcal{O}caligraphic\_O such that 𝒪⊆𝒪p𝒪subscript𝒪𝑝\mathcal{O}\subseteq\mathcal{O}\_{p}caligraphic\_O ⊆ caligraphic\_O start\_POSTSUBSCRIPT italic\_p end\_POSTSUBSCRIPT, |𝒪|≤k𝒪𝑘|\mathcal{O}|\leq k| caligraphic\_O | ≤ italic\_k and L(𝒪)≤ℓ𝐿𝒪normal-ℓL(\mathcal{O})\leq\ellitalic\_L ( caligraphic\_O ) ≤ roman\_ℓ. ‘No’ otherwise.
We prove the theorem by reduction from the decision version of the set-cover problem—known to be NP-complete—to OI-DEC.
The set-cover problem is defined as follows.
Definition 8 (SetCover-DEC):
Given a set of elements 𝒰𝒰\mathcal{U}caligraphic\_U, a set of subsets 𝒳={X⊆𝒰}𝒳𝑋𝒰\mathcal{X}=\{X\subseteq\mathcal{U}\}caligraphic\_X = { italic\_X ⊆ caligraphic\_U }, and an integer k𝑘kitalic\_k, return ‘Yes’ if there exists a cover 𝒞⊆𝒳𝒞𝒳\mathcal{C}\subseteq\mathcal{X}caligraphic\_C ⊆ caligraphic\_X such that ⋃X∈𝒞X=𝒰subscript𝑋𝒞𝑋𝒰\bigcup\_{X\in\mathcal{C}}X=\mathcal{U}⋃ start\_POSTSUBSCRIPT italic\_X ∈ caligraphic\_C end\_POSTSUBSCRIPT italic\_X = caligraphic\_U and |𝒞|≤k𝒞𝑘|\mathcal{C}|\leq k| caligraphic\_C | ≤ italic\_k. ‘No’ otherwise.
If there is some u∈𝒰𝑢𝒰u\in\mathcal{U}italic\_u ∈ caligraphic\_U that is not included in at least one of the subsets X𝑋Xitalic\_X, then the answer is ‘No’.
Assuming otherwise, we construct an instance of a shortest path problem (a special case of an MDP problem) as follows (Figure [2](#S3.F2 "Figure 2 ‣ Proof. ‣ 3 Complexity Results")).
There are four types of states in the MDP: (1) ui∈𝒰subscript𝑢𝑖𝒰u\_{i}\in\mathcal{U}italic\_u start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_U represents one of the elements in 𝒰𝒰\mathcal{U}caligraphic\_U, (2) Xi∈𝒳subscript𝑋𝑖𝒳X\_{i}\in\mathcal{X}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_X represents one of the subsets in 𝒳𝒳\mathcal{X}caligraphic\_X, (3) Xi′∈𝒳′subscriptsuperscript𝑋′𝑖superscript𝒳′X^{\prime}\_{i}\in\mathcal{X}^{\prime}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT: we make a copy for every state Xi∈𝒳subscript𝑋𝑖𝒳X\_{i}\in\mathcal{X}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_X and call them Xi′subscriptsuperscript𝑋′𝑖X^{\prime}\_{i}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT, (4) a goal state g𝑔gitalic\_g.
Thus, the state set is 𝒰∪𝒳∪𝒳′∪{g}𝒰𝒳superscript𝒳′𝑔\mathcal{U}\cup\mathcal{X}\cup\mathcal{X}^{\prime}\cup\{g\}caligraphic\_U ∪ caligraphic\_X ∪ caligraphic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ∪ { italic\_g }.
We build edges between states as follows:
(1) e(u,X)∈E𝑒𝑢𝑋𝐸e(u,X)\in Eitalic\_e ( italic\_u , italic\_X ) ∈ italic\_E iff u∈X𝑢𝑋u\in Xitalic\_u ∈ italic\_X: For u∈𝒰𝑢𝒰u\in\mathcal{U}italic\_u ∈ caligraphic\_U and X∈𝒳𝑋𝒳X\in\mathcal{X}italic\_X ∈ caligraphic\_X, there is an edge between u𝑢uitalic\_u and X𝑋Xitalic\_X.
(2) ∀Xi∈𝒳for-allsubscript𝑋𝑖𝒳\forall X\_{i}\in\mathcal{X}∀ italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_X, e(Xi,Xi′)∈E𝑒subscript𝑋𝑖subscriptsuperscript𝑋′𝑖𝐸e(X\_{i},X^{\prime}\_{i})\in Eitalic\_e ( italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT , italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ) ∈ italic\_E: For every Xi∈𝒳subscript𝑋𝑖𝒳X\_{i}\in\mathcal{X}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT ∈ caligraphic\_X, we have an edge from Xisubscript𝑋𝑖X\_{i}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT to Xi′subscriptsuperscript𝑋′𝑖X^{\prime}\_{i}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT. (3) ∀e(X′,g)∈Efor-all𝑒superscript𝑋′𝑔𝐸\forall e(X^{\prime},g)\in E∀ italic\_e ( italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT , italic\_g ) ∈ italic\_E: for every X′∈𝒳i′superscript𝑋′subscriptsuperscript𝒳′𝑖X^{\prime}\in\mathcal{X}^{\prime}\_{i}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ∈ caligraphic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT we have an edge from Xisubscript𝑋𝑖X\_{i}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT to the goal g𝑔gitalic\_g.
This construction can be done in polynomial time.
Let M𝑀Mitalic\_M be the MDP constructed in this way.
We show that SetCover(𝒰,𝒳,k𝒰𝒳𝑘\mathcal{U},\mathcal{X},kcaligraphic\_U , caligraphic\_X , italic\_k) = OI-DEC(M,V0=0,k,2formulae-sequence𝑀subscript𝑉0
0𝑘2M,V\_{0}=0,k,2italic\_M , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT = 0 , italic\_k , 2).
Note that by construction every state Xisubscript𝑋𝑖X\_{i}italic\_X start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT, Xi′subscriptsuperscript𝑋′𝑖X^{\prime}\_{i}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT italic\_i end\_POSTSUBSCRIPT, and g𝑔gitalic\_g converges to its optimal value within 2 iterations as it reaches the goal state g𝑔gitalic\_g within 2 steps. A state
u∈𝒰𝑢𝒰u\in\mathcal{U}italic\_u ∈ caligraphic\_U converges within 2 steps if and only if there exists a point option (a) from X𝑋Xitalic\_X to g𝑔gitalic\_g where u∈X𝑢𝑋u\in Xitalic\_u ∈ italic\_X, (b) from u𝑢uitalic\_u to X′superscript𝑋′X^{\prime}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT where u∈X𝑢𝑋u\in Xitalic\_u ∈ italic\_X, or (c) from u𝑢uitalic\_u to g𝑔gitalic\_g. For options of type (b) and (c), we can find an option of type (a) that makes u𝑢uitalic\_u converge within 2 steps by setting the initial state of the option to ℐo=Xsubscriptℐ𝑜𝑋\mathcal{I}\_{o}=Xcaligraphic\_I start\_POSTSUBSCRIPT italic\_o end\_POSTSUBSCRIPT = italic\_X, where u∈X𝑢𝑋u\in Xitalic\_u ∈ italic\_X, and the termination state to βo=gsubscript𝛽𝑜𝑔\beta\_{o}=gitalic\_β start\_POSTSUBSCRIPT italic\_o end\_POSTSUBSCRIPT = italic\_g.
Let 𝒪𝒪\mathcal{O}caligraphic\_O be the solution of OI-DEC(M,k,2𝑀𝑘2M,k,2italic\_M , italic\_k , 2).
If there exists an option of type (b) or (c), we can swap them with an option of type (a) and still maintain a solution.
Let 𝒞𝒞\mathcal{C}caligraphic\_C be a set of initial states of each option in 𝒪𝒪\mathcal{O}caligraphic\_O (𝒞={ℐo|o∈𝒪}𝒞conditional-setsubscriptℐ𝑜𝑜𝒪\mathcal{C}=\{\mathcal{I}\_{o}|o\in\mathcal{O}\}caligraphic\_C = { caligraphic\_I start\_POSTSUBSCRIPT italic\_o end\_POSTSUBSCRIPT | italic\_o ∈ caligraphic\_O }).
This construction exactly matches the solution of the SetCover-DEC.
u1subscript𝑢1u\_{1}italic\_u start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPTu2subscript𝑢2u\_{2}italic\_u start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPTu3subscript𝑢3u\_{3}italic\_u start\_POSTSUBSCRIPT 3 end\_POSTSUBSCRIPTu4subscript𝑢4u\_{4}italic\_u start\_POSTSUBSCRIPT 4 end\_POSTSUBSCRIPTu5subscript𝑢5u\_{5}italic\_u start\_POSTSUBSCRIPT 5 end\_POSTSUBSCRIPTX1subscript𝑋1X\_{1}italic\_X start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPTX2subscript𝑋2X\_{2}italic\_X start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPTX1′subscriptsuperscript𝑋′1X^{\prime}\_{1}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPTX2′subscriptsuperscript𝑋′2X^{\prime}\_{2}italic\_X start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPTg𝑔gitalic\_g
Figure 2: Reduction from SetCover-DEC to OI-DEC. The example shows the reduction from an instance of SetCover-DEC which asks if we can pick two subsets from 𝒳={X1,X2}𝒳subscript𝑋1subscript𝑋2\mathcal{X}=\{X\_{1},X\_{2}\}caligraphic\_X = { italic\_X start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT , italic\_X start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT } where X1={1,2,3},X2={3,4,5}formulae-sequencesubscript𝑋1123subscript𝑋2345X\_{1}=\{1,2,3\},X\_{2}=\{3,4,5\}italic\_X start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT = { 1 , 2 , 3 } , italic\_X start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT = { 3 , 4 , 5 } to cover all elements 𝒰={1,2,3,4,5}𝒰12345\mathcal{U}=\{1,2,3,4,5\}caligraphic\_U = { 1 , 2 , 3 , 4 , 5 }. The SetCover-DEC can be reduced to an instance of OI-DEC where the question is whether the MDP can be solved with 2 iterations of VI by adding at most two point options. The answer of OI-DEC is ‘Yes’ (adding point options from X1subscript𝑋1X\_{1}italic\_X start\_POSTSUBSCRIPT 1 end\_POSTSUBSCRIPT and X2subscript𝑋2X\_{2}italic\_X start\_POSTSUBSCRIPT 2 end\_POSTSUBSCRIPT to g𝑔gitalic\_g will solve the problem), thus the answer of the SetCover-DEC is ‘Yes’. Here the set of initial states corresponds to the cover for the SetCover-DEC.
∎
###
3.1 Generalizations of MOMI and MIMO
A natural question is whether Theorem [1](#Thmtheorem1 "Theorem 1. ‣ 3 Complexity Results") extends to more general option-construction settings. We consider two possible extensions,
which we believe offer significant coverage of finding optimal options for planning in general.
We first consider the case where the options are not necessarily point options. There is little sense in considering MOMI where one can choose any option since clearly the best option is the option whose policy is the optimal policy. Thus, using the space of all options 𝒪allsubscript𝒪𝑎𝑙𝑙\mathcal{O}\_{all}caligraphic\_O start\_POSTSUBSCRIPT italic\_a italic\_l italic\_l end\_POSTSUBSCRIPT we generalize MOMI as follows:
Definition 9 (MOMIgen𝑔𝑒𝑛{}\_{gen}start\_FLOATSUBSCRIPT italic\_g italic\_e italic\_n end\_FLOATSUBSCRIPT):
Given an MDP M𝑀Mitalic\_M, a non-negative real-value ϵitalic-ϵ\epsilonitalic\_ϵ, an initial value function V0subscript𝑉0V\_{0}italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT, 𝒪′⊆𝒪allsuperscript𝒪normal-′subscript𝒪𝑎𝑙𝑙\mathcal{O^{\prime}}\subseteq\mathcal{O}\_{all}caligraphic\_O start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT ⊆ caligraphic\_O start\_POSTSUBSCRIPT italic\_a italic\_l italic\_l end\_POSTSUBSCRIPT, and an integer ℓnormal-ℓ\ellroman\_ℓ, return 𝒪𝒪\mathcal{O}caligraphic\_O minimizing |𝒪|𝒪|\mathcal{O}|| caligraphic\_O | subject to Lϵ,V0(𝒪)≤ℓsubscript𝐿italic-ϵsubscript𝑉0𝒪normal-ℓL\_{\epsilon,V\_{0}}(\mathcal{O})\leq\ellitalic\_L start\_POSTSUBSCRIPT italic\_ϵ , italic\_V start\_POSTSUBSCRIPT 0 end\_POSTSUBSCRIPT end\_POSTSUBSCRIPT ( caligraphic\_O ) ≤ roman\_ℓ and 𝒪⊆𝒪′𝒪superscript𝒪normal-′\mathcal{O}\subseteq\mathcal{O^{\prime}}caligraphic\_O ⊆ caligraphic\_O start\_POSTSUPERSCRIPT ′ end\_POSTSUPERSCRIPT.
######
Theorem 2.
MOMIgen𝑔𝑒𝑛{}\_{gen}start\_FLOATSUBSCRIPT italic\_g italic\_e italic\_n end\_FLOATSUBSCRIPT and MIMOgen𝑔𝑒𝑛{}\_{gen}start\_FLOATSUBSCRIPT italic\_g italic\_e italic\_n end\_FLOATSUBSCRIPT are NP-hard.
The proof follows from the fact that MOMIgen𝑔𝑒𝑛{}\_{gen}start\_FLOATSUBSCRIPT italic\_g italic\_e italic\_n end\_FLOATSUB
| -1
|
Unrelated
| false
|
<urn:uuid:9e90b6ba-0e0c-45a2-bb3a-74e7c1edc8cb>
|
dclm-dedup-25B-ai-scifi-docs | http://friendofsophia.blogspot.com/
|
Rannm Thawts Ten
Random thoughts.
• People keep citing Clarke's Third Law (the "sufficiently advanced technology" one) as if it's a real scientific principle. It's not. It's Clarke demonstrating that island savages are prone to Cargo Cultism, no matter what island they live on. Because magic doesn't have to take thermodynamics into account; technology does. Go look up how hard it is just to make things float, via technology; now consider how easy it generally is for wizards (moving objects up to 5 pounds is a zeroth-level spell, in D&D).
This is not only why hover-tanks and "nanomachines are magic" plots are stupid, it's also why shows like Gate: Thus the JSDF Fought There are stupid. A wizard who can make portals isn't going to be impressed by physics, except in the sense that you're impressed by things some disabled people can do; they can already do things our physics says are probably impossible and almost certainly practically impossible even if they can technically happen.
And a "great red dragon" in a blatantly D&D-based setting isn't going to be impressed by your tank shells, son; it's wholly immune to the half that's fire damage and little if any of the half that's just regular damage is going to get through its damage reduction. Those five or so HP of damage you might do are going to piss it off, though—it'll probably take about six seconds to land, dig open the hatch, and turn the crew (and upholstery) into a fine coating of white ash inside the tank. Maybe you don't know how many attacks a dragon gets in a full-attack action?
• The reason critics praise "subversion," even when it's manifestly moronic, and will defend even mean-spirited, incoherent dreck like Star Wars: The Last One Anybody Will See in Theaters, has little to do with politics or being adherents of post-structuralist or postmodern ideologies, and much to do with the fact critics are unhappy people, basically damned while still alive. You see, to be a critic is to do something that real humans do for fun, as your job.
Film critics, for example, go to see every movie, whether they want to or not. They see far more movies than anyone else. Hence why they habitually mistake all tropes for clichés (the fact they don't know tropes from clichés is why assertions that they're some kind of ideologues are doubtful: they would need real educations for that). Hence also why they will snap up anything novel, no matter how mean-spirited or half-assed. They're dead inside, and novelty is the only thing that makes them feel anything.
• It is 100% fair to call Thundercats Roar badly-drawn crap. Ditto Steven Universe, though its bad art is the least of that show's problems. But it is not fair to call that art-style "CalArts"; that term, as a form of abuse, was actually coined by John Kricfalusi, the talentless psychopath behind Ren and Stimpy. And he actually applied it to the usual form of Disney animation. Which he presumably didn't like because, unlike his art, it doesn't look like an unsolicited dick pic. (I'm not really picking that analogy at random.) Also the hacks behind Steven Universe went to the School of Visual Arts in New York.
The art-style of Steven Universe and Thundercats Roar, aside from being much closer to Kricfalusi's art style than to the one the mongrel was attacking, really ought to be called "Tumblr Arts", because that's the place you'll see it most. Remember that "let people enjoy things" comic that's the only defense that people with no taste can make of the trash they're into? That art style. Now, admittedly, good shows have been made in a similar style—Gravity Falls and Star vs. the Forces of Evil, for instance. But those shows are made by people who know what they're doing, unlike Steven Universe or Thundercats Roar.
• My Common Tongue has an agreement system somewhat similar to Uto-Aztecan or Bantu. Because they're prefixes rather than unbound morphemes, it's kinda hard to use possessives predicatively ("this dog is mine" vs. "this is my dog") in Uto-Aztecan languages, and predicative possessives are important in a particular type of phrasing that I like. Tribute must be paid to the greatest fantasy currently being done in English, as once 'twas paid unto Tolkien and Howard and Vance. But there is an equivalent, in Nahuatl: you basically say "O [...] that I have" rather than "O [...] mine".
The power of undeath behind the nightshades, by the bye, talks in trochaic heptameter. I'm not sure how that actually works in the Common Tongue; I also haven't really worked out how the beast-totem chants being in trochaic tetrameter ("Kalevala meter") works. I should probably give their poetry more kennings and parallelisms, which A, work largely independent of language (one of the reasons the Bible is such a great work of literature is its parallelisms usually translate well—in which you may certainly see the hand of God if you choose), and B, are the two features that define Nahuatl poetry.
• It occurs to me, that theme I like about how individualism and collectivism are both really bad for civilization, and are fundamentally errors with regard to the Problem of Universals, is also kinda similar to the existentialist concept of "bad faith". Except that existentialism mainly starts from the ethics end and I start at the epistemology/metaphysics end. Existentialist epistemology is generally pretty vague, if not actually incoherent; it thus tends to be too easily corrupted into Postmodernism and Social Constructionism, where all truth is reduced to power-relationships—or as those schools' most consistent adherents know the concept, the Sword Logic.
• I don't understand people's inability to be pleased. There are mongrels claiming that the writers of Halo 5 didn't know who the game is about (you're actually fighting logic if you just deny that it's the best game in the series except ODST and maybe Reach). I admit I automatically award significant bonus-points just for not involving the Flood, who as I've said turn a top-notch shooter into third-rate survival horror, and for having been actually playtested (not like that's the only reason Halo 3 is better than Halo 2, but it's a big one...though admittedly Halo 3 does have the level "Cortana").
Others of these beasts of the field will claim that Destiny 2 is worse than the first one in every way, which is actually the opposite of true. The second has a better inventory system, a better interface, better loot, and public events are much easier to participate in. Yes, Warmind was kinda lackluster, and while Curse of Osiris isn't terrible it could've stood to be longer and go more places (there was apparently some funny business with the experience calculation, which is an issue of the game as a product but not of the game as a "text").
However, it's not like The Dark Below was particularly brilliant, and I personally don't give a damn about Rise of Iron beyond its resolution of the Fallen plotline making Destiny 2 make sense. Hell The Taken King is near-universally regarded as the best expansion of the first game (I don't know how so many people can misspell "House of Wolves" like that), and that was when your character became a mime, for no apparent reason. Also the Taken show up in various areas before your character has actually encountered them in the game's story (which you'll note they don't, in the second game).
• Reading a lot of tie-in novels lately; there's a summer-reading thing at my local library. I find I like tie-in fantasy more because I don't have to sit while Sandon Branderson or somebody lucubrates on forty-three different kinds of metamorphic rock and how each affects the color of your astral cord when you mix your astral-projection potion in a mortar made of it.
One thing I noticed is that not only are the Warhammer Fantasy novels less pointlessly grimdark than ASoIaF (despite being the people who literally invented it), they're actually less pointlessly grimdark than the Pathfinder ones. Ain't even passing references to people being raped by ogres (or "greenskins"), in Warhammer. It's basically impossible for Pathfinder to mention ogres without that coming up.
I'm really looking forward to Kingmaker, but I can't escape the worry that I'm going to be subjected to something out of a tenth-grade creative-writing club-member's attempt to be edgy.
• Noticed something watching E3: people are actually praising "gritty" environments. Um...what? Every game has "gritty" environments, and basically has ever since the hardware was up to displaying that many objects on-screen. Actually what they should be praising is the few games where everything isn't bombed-out hovels plagued by nuclear mutants. At least Destiny is the ruins of a bunch of space-colonies, but would it seriously kill you people to have a video game where people don't all have gravel-pits in the middle of the living room?
Sure, the occasional bombed-out building makes sense, in a shooter or war-game, even an RPG or open-world. Every building being a bombed-out shell? No. Halo 5, especially in the Sanghelios levels, hit a nice balance between clean modern buildings, ancient ruins, and bombed wreckage, and when the Guardian started breaking things in Sunaion it actually meant something. I suppose this is just a broader thing about how post-apocalyptic settings are fundamentally lazy; even in Destiny the "wreckage of
the Golden Age" thing is the weakest part of the setting.
• Tangentially related to the tie-ins thing, it is utterly inexplicable to me that 40K is more popular than Fantasy Battle. The black-and-gray morality of WHFB was Flanderized into evil-vs.-evil; the Empire that could maintain cordial relations with elves and dwarfs became genocidal totalitarians. The one time science fiction (in the very broad sense of "set in space in the future") does better than (traditional) fantasy, and it's the markedly inferior product!
Playing with Fantasy VIII
Fantasy game thoughts.
• I'm not tired any more, so I did the number-crunching. A dragon of the dimensions of a river otter, but 120 feet long, and only as dense as a bird so massing 69.6 (short) tons, with a wingspan of 108 feet, would, assuming its neck and tail include feathers to act as lifting-area and it is, thus, basically kite-shaped (but leaving off say 10% of the length, for the head itself—a square kite, basically, although the back is longer and the front is shorter), have a wing-area of 5,832 square feet and a wing-loading of 116.5 kilograms per square meter. That results in a takeoff speed of 88.2 miles per hour. The wings are also not just triangles, they're shaped more like a bird-wing, but that's the net total area.
I wonder if the really big dragons run down mountain slopes to get up to speed more quickly. For the gold-dragon sized ones, the younger age-categories would weigh only 35.5%, 9.6%, 1.7%, and 0.23% as much, at the Gargantuan, Huge, Large, and Medium age-categories respectively, and yet their wing area would be only 50.2%, 21%, 6.7%, and 1.8% the area, so the wing-loading goes down drastically. (Small and Tiny, found in smaller types of dragons at young age-categories, are 0.03% and 0.004% as heavy and have wings with 0.4% and 0.1% the area.) Actually, let me crunch the takeoff speeds for 'em all: Gargantuan, 74.2 mph; Huge, 59.7 mph; Large, 44.8 mph; Medium, 32.2 mph; Small, 22.8 mph; Tiny, 16.1 mph. I.e. the large one just has to move as fast as a fast horse to take off.
You can actually move something built like an otter pretty quickly; rabbits, after all, have a similar body-plan.
• I'd been struggling with my Fiendish/Celestial/Primordial/(Aklo) language. There isn't enough of a corpus of Valarin, Black Speech, or whatever you want to call the Cthulhu gibberish (it's not Aklo, I'll tell you that for free) to easily make a language based on any of them. (Though they did do a pretty good job with the "Faceless" language in WoW, but like I said, basing the phonics on Cthulhu gibberish was a chore to pronounce even for me.)
I eventually buckled down, bit the bullet, and just overhauled the grammar to the point of actual usefulness, but along the way I toyed with just declaring that there is no such language, as we think of language. I had two rationales (or rationalizations) for that. One, they're divine beings, so glossolalia (speaking in tongues) as their mode of expression makes a kind of sense; and two, my setting is partly based on Native American ideas, with only Old World material culture. The Navajo gods are defined as unable to speak. (Yes, even "Talking God"; he metaphorically speaks for them, as their leader.)
The way that would have worked, if I hadn't eventually gotten down to business, is that anyone who speaks the divine/extraplanar language would be able to understand anyone else speaking it, as if they spoke the same language, but really they're just babbling glossolalia at each other.
• One thing I decided along the way is that all the "outsiders", not just the fiendish ones, have names in Primordial, but the gods prefer the names in the languages their mortal children have given them. Whereas the fiends prefer to be called on by their original names, if not in their own languages, because they view mortals as livestock, not even pets let alone children.
Now, of course, I have to come up with a system for creating names for fiends, which system I can also use for the courtesy-names of mortal witches. (Actually maybe just human witches, the dark-elf and black-dwarf witches don't worship fiends like human ones do, they worship gods that happen to be hostile to the other gods. Goblins and orcs don't have witches.)
Think maybe the fiends' names will have a third element, though, to keep the talking pond-scum in its place.
• I think I can get a reasonable lift out of the Pathfinder Ultimate Combat airship, with a steam-filled envelope (I draw the line at letting a fantasy society have helium, and hydrogen is suicide). Steam has about 61% (actually 20/33) as much lift as helium, so you need 65% more volume; medieval ships the size of their airship's gondola, 20 feet by 60 feet, typically have displacements of 20 to 30 tons, plus 30 tons of cargo. A helium-envelope to lift 55 tons would be 1,581,715.41 cubic feet, so a steam one is 2,609,830.43 cubic feet. Assuming the same proportions as its gondola, that means an envelope 355.32 feet long and 118.44 feet wide (and tall).
Of course, we're glossing over the fact it's really hard to contain superheated steam safely. Handwave it with "magically treated" material, and so on. I think the steam is magically generated somehow (fire and water elementals in some kind of ethically questionable harness?). The "magical engine" in the vehicle description is vague; my gut instinct, of course, is that it should be a pretty chair that eats the day's spellcasting of a spellcaster who sits in it, but that doesn't really match the actual description (also it's probably copyright infringement). I picture it as a big stone pillar with runes that both indicate and let you control your altitude and speed.
• One thing the Elder Scrolls setting does remarkably well, but that most of the audience probably missed, is Gnostic twaddle (though really if you're not familiar with Gnostic twaddle it probably speaks to your good judgment). Read, for example, The 36 Lessons of Vivec, and then read something like the Gospel of Judas: the exact same type of self-satisfied, self-important bafflegab, dressing up deeply shallow pseudo-philosophy in big, impressive-sounding buzzwords. I don't mean this as a criticism; it's a fascinating way to develop a setting, by giving its mystics authentic esoteric gobbledygook. (Also, as I think I've said before, it's nice that all those people with comparative religion degrees are finding work.)
• Decided that, just as my setting only has one kind of fiend, it only has the angel-type celestials. Other than that there's the elementals. I might keep the guardinals agathions, eladrin azatas, and inevitables as servitors of the human, elf, and dwarf gods. But then again maybe not, since I can't really find anything appropriate to use for servitors of the gnome gods. (The Pathfinder "Dimension of Dreams" is sorely lacking in anything one might use that way, practically everything you meet there being straight-up evil instead of merely incredibly dangerous through no fault of their own, as would make sense in a world run on "dream logic".)
I was starting to think I'd use a lot more fey than I'd thought I would—fauns but not satyrs; dryads, hamadryads, nereids, and oceanids but not nymphs; atomies and pixies but not the others—but no, I think I'll just have things like genies count as "fey" for purposes like a druid's Resist Nature's Lure ability. The last straw was how Pathfinder conflates rusalka with bludička (the ara-mitama of the rusalka), which completely screws up the ending of the opera. Also vodyanoi certainly do not "resemble humanoid salamanders". They're water goblins. Their theme-song is even often called that, in English.
Basically the whole edifice of the "fey" creature type, in a world with elves and dwarves (or goblins), was weird from the get-go; and Pathfinder trying to make the gnomes more a part of it than the others was even more bizarre. Elves, dwarves, and goblins actually are fairies (except in Germanic languages instead of Romance ones), whereas gnomes are elemental spirits from an alchemist's cosmological speculations. (Also though seriously the other word Paracelsus used for them, in his Latin notes? Pygmaeus…the Greco-Latin for "dwarf"! What a man whose real name was Philipp Bombast von Hohenheim might mean by "dwarf" is left as an exercise for the student.)
Basically, what D&D calls a gnome really should've been called a brownie, since the actual gnomes were just dwarves. Yes I realize "jinn" is pretty much just "fairy" in Arabic. Even I'm not that much of a stickler, though.
• People complain about feasting in fantasy novels. I'm not sure why; probably the stupid idea that what does not directly advance the "plot" is bad, never mind a well-written feast actually advances plot too quickly, if anything. I can see complaining about a paper-thin Ren Faire cliché storm feast (giant turkey-legs, huge carcasses being spit-roasted), but I mean, can you find Japan on a map? Or any other Pacific island? Heard of the Tlingit? And, yes, the Norse? Feasts are a huge deal, anthropologically; they cement relationships and allow the elite to display their power without having to kill anyone. Gifts are given at feasts, and songs are sung. If you can't figure out how these things are a convenience to a fantasy story, you have no business reading them, let alone writing them.
I'd actually like to see feasts in fantasy games—have that be where you find out the ancient prophecy you're supposed to fulfill, or where you're gifted your plot-significant weapon, from the largess of a mighty chief. Oh, but they'd be boring to sit through? Most of the Thieves' Guild questline in Skyrim consists of standing around while NPCs talk; a feast would at least establish setting, even if you stupidly decided not to have them be where key story-development occurs. You should get a feast every time you become a thane, and maybe have a skald sing something that gives you a tip for fighting Alduin, make the last fight easier. That would certainly be better than entire Mephala and Boethiah questlines that wound up being cut anyway.
• Decided that the giants in my setting are from the gas giants in the system (Neptune- and Uranus-type gas giants, with solid cores); they had to abandon their worlds at the same time the elves and dwarves abandoned the moons. Decided that wood and frost giants have the proportions of elves, while stone and fire have the proportions of dwarves and hill have the proportions of humans (this results in a 12-foot-6-inch fire or stone giant to a 15-foot wood or frost giant, and a 13-foot-9-inch hill giant). Each group of proportions is from a different gas giant.
Also decided that the fire and frost giants are the giant equivalents of orcs or black dwarves and goblins or dark elves, respectively, changed by trafficking with a dark power (an outcast member of their pantheon). My wood and stone giants have cold and fire resist 5, while the "changed" equivalents have full immunity to the energy-type in question. The hill giants were all changed, the way the frost and fire giants were, but mine are a bit smarter than the ones in the core rules (say Int 8 or 9 instead of 6). They're giant humans, basically.
Might change it so giants advance by class-levels like other humanoids, and have all the hill giants be barbarians while the others are mostly warriors.
Sierra Foxtrot 11
SF thoughts.
• Was doing some research on quantum computing. Turns out, while quantum processing is hugely advantageous, storage still pretty much has to be "classical" (here meaning just "not quantum"), certainly if you ever want to copy things; but quantum computing would tend to work with much bigger memories. The solution is apparently to find some way to store your data in three dimensions. Some people recommend DNA, but that seems really suspect, and (given how much we still don't understand about DNA, and how complicated it is even when we do understand it) prone to all kinds of bugs. I think a better method would be so-called "holographic data storage".
• I have, like most thinking people, only what tolerance for "dark matter is magic" is strictly necessary to keep watching shows like The Flash. (A show that, like Arrow, has a bigger problem, namely that they're clearly having Hal "I'm such a bad boyfriend my girlfriend became a supervillain" Jordan write their romance subplots.) The thing about dark matter is it doesn't interact with normal matter, except by gravity, so while it has very weird properties, they probably aren't very useful. Better that than "nanomachines are magic", though, I suppose.
But if you must have something relating to dark matter be related to your mystical foofaraw, at least dress it up a bit. Destiny, for example, although they have dark matter be an indicator of the reality-warping powers of the Darkness (no idea if there's some similar indicator of the powers of the Light), at least say "sterile neutrinos", which you have to look up to know they're associated with dark matter. (Regular, "active" neutrinos interact via the weak force, only.) And no, SIVA isn't magic nanomachines, it can only kind of infect Ghosts, for a reason—in that setting, "magic nanomachines" would be meant literally.
• I think it's ironic, since the Dune series was written as an attack on the idea of hero-worship, that the only parts of it anyone remembers are the parts that would lead to hero-worship. (Well, I also often quote Harkonnen's line about "Never trust a traitor, even one you created yourself.") It's like François Truffaut's famous line, "...Some films claim to be antiwar, but I don't think I've really seen an antiwar film. Every film about war ends up being pro-war."
It's also ironic that Herbert actually listed the Jesuits as one of the great tyrannical systems of history, in one of the sequels. Um...what? No like seriously what? The Jesuits were suppressed in 1773 at the behest of the European empires, because they didn't like all these priests gumming up their tyrannical systems. Jesuit missionaries made a nuisance of themselves, advocating for the natives and building communities that allowed the natives to be self-sufficient and independent of the colonial governments.
Was Herbert maybe thinking of the Dominicans? At least that would make sense with the Spanish Inquisition, even though the Inquisition was the mildest Early Modern ideological court-system. (Of course, because of the Inquisition, Spain had basically no witch-hunts. Unlike most of the people who pretend to be so shocked by the Inquisition doing much milder things than all their own courts were doing outside their witch-hunts.)
• Remember how I was wondering how termites replace their queens if the old one dies, when all the candidates would be the daughters of the old queen and thus also of the king? Turns out, termite queens can parthenogenetically produce clones of themselves to replace them; some of them are on hand in any given colony at any one time, in case the old queen dies.
No idea how you get new kings if the old one dies, though (termites aren't Hymenoptera, their eggs require fertilization each time they're produced, like the rest of us do it). Maybe a queen dies too when her mate does, and then one of her clones does a mating flight with a king from outside, that isn't the son of the old queen.
Turns out that termites are in the order Blattodea, same as cockroaches, not just related to it (Isoptera, their old order, turns out not to exist). They have a bunch of behaviors in common, like pheromone trails and kin-recognition. Of course, cockroaches' aversion to light doesn't extend to all being blind, as non-alate termites typically are. The order's closest relative is the one mantises are in, Mantodea.
• I'm curious, people who subscribe to the "stronger" climate-change predictions (the milder, likelier ones are less likely to show up in science fiction, as well as being harder to milk political capital out of): why do you keep saying we're going to see droughts?
Cold is dry; in a Glacial Maximum, most of Africa and significant chunks of Eurasia and the Americas are uninhabitable desert. Heat is wet, because less of the water is locked up in glaciers—even in warmer phases of this glaciation period, large portions of the Sahara are forest.
If your conception of climate change involves global cooling, e.g. us accidentally skewing things back toward a glacial maximum (or even just a higher level of glaciation), then of course this remonstration is not directed at you.
• You've probably come across the idea of the "motherhood statement", and the idea that good science fiction comes from "burning the motherhood statement" (it's usually mentioned in the "standard" version of the Turkey City Lexicon, for instance). Which I think just proves a significant portion of the science fiction fandom actually doesn't give a damn about science, except as window-dressing for their actually Gnostic views. Because, I mean, are we supposed to just deny evolutionary theory? Even Heinlein knows that what you're "for", biologically speaking, is reproduction—"motherhood"—and nearly everything else is in service to that. If you're more unrealistic and Gnostic in your views of human sexuality and families than Heinlein, you have a problem.
• Apparently rats laugh when they're tickled, and their ears droop and turn pink when they're happy. The really interesting thing is that when they laugh, we can't hear it—it's too high-pitched. (Many rodent vocalizations are, that's why things that hunt them, like foxes and cats, have such good high-frequency hearing.)
Another thing this presumably means is that blushing and laughter either predate the split between Euarchonta (tree-shrews, colugos, and primates) and Glires (rodents and lagomorphs), or else independently evolved in both. My money is maybe on the first one? Though I wonder what purpose flushing with blood when emotional serves in a rodent: the ability to see red only evolved with the simians (though the evolution of color vision is complicated, between Old World and New World monkeys).
• Speaking of the unusual ability to see long wavelengths of light, vampire bats and pit-vipers independently evolved infrared vision that uses thermoreceptors near their noses and connects to their optic nerves. A lot of the brain-structures involved are even analogous, despite the last common ancestor of bats and snakes being a basal reptiliomorph from about a third of a billion years ago.
• Something people are apparently realizing is unrealistic in a lot of science fiction, is the Gattaca-type stuff where society's "haves" have designer children and the "have-nots" don't (and which Gundam SEED should've been about, but wasn't, because that show is stupid). Now, it is true that realistically it won't make enough of a difference, because genetic enhancement is still partly a crapshoot if you don't utterly reorganize everything else in the subject's life to also work toward your desired result. But the assertion of unrealism is itself unrealistic, for one reason.
Namely, just because you're not remotely guaranteed to get the super kid you want, won't stop people from trying. This is a species that practiced trepanning, footbinding, and tightlacing, do you think it's going to let a little thing like "it isn't actually all that likely to work" stop it?
I'd actually like plots with yuppie-scum whining about all the money they wasted to make their kid a genetic shoo-in for the Ivy League, and then it turns out the only League their kid cares about is the "of Legends" variety. But I don't think people (certainly not people who are published by "traditional", i.e. gatekept, publishing houses) are quite ready to face that specific social commentary; hits a bit close to home given where and by whom the publishing industry is run. (Of course, given the median Harvard grade is A- and the mode is A, the Ivy League has other issues...)
Playing with Fantasy VII
Fantasy RPG thoughts.
• I'd gotten rid of trolls in my campaign, but then I got to thinking, maybe make 'em like a yeti-sasquatch-abominable snowman thing? Could just call 'em "abominations". "Snow abomination" = frost troll, maybe. Apparently the main Nepali name, himamaanav, just means "snow man" (they may well call the child's ice-sculpture something else, like "snow Bodhidharma"); one of the Tibetan names, meaning "wild man", is "mi-go". How exactly Lovecraft managed to equate the two is a question for the ages.
I never much cared for the troll social behavior as presented starting I think in 3rd Edition (at least I don't remember any mention of troll matriarchs back in 2nd). Think mine'll be more like certain reptiles, which lay their eggs and then their young are on their own. Nobody ever said trolls don't lay eggs, and none of the rest of their behavior seems to go with K-selection. How do trolls without "adult supervision" not overrun an ecosystem? Young ones can get eaten by big predators—stomach acid stops their regeneration.
• It occurs to me that having a glowing iris but a dark pupil, combined with sclera having the same appearance as the iris—the norm for non-human animals—would give you the "whole eye glows" glowing eyes seen in Warcraft. Especially if you also have it so the pupil completely seals shut and the iris and sclera are the receptor for whatever energy darkvision perceives? Maybe darkvision is something like a parietal eye or the heat-sensing "pits" in a pit-viper or a vampire bat, but built into the outer surface of the eye rather than a separate organ. And sensing some weird magical energy (or maybe radar, which is honestly the thing most like how darkvision behaves, but if you can see your surroundings by passive radar on a planet's surface, you live in a very odd environment).
My fiends also all have three eyes, and the third is the one that gives them see-in-darkness and an at-will deathwatch ability. I was also toying with doing something weird with dragon eyes. One that was basically automatic was comparable visual acuity to a bird of prey (de rigueur for a flying predator—and presumably pretty easy to accomplish when your eyeball is the size of a shot-put ball), but then I thought maybe two pupils so they can do parallax-based depth perception from a single eye? But then, even better, was monocular depth-perception via "corneal accommodation", like a chameleon. After all, sub in the breath-weapon for the chameleon tongue and you've got a dragon. Presumably they don't put the eye on a turret the way chameleons do (since their head is a lot more mobile than a chameleon's).
• Decided to use a river otter, specifically the giant river otter, as the model for dragon anatomy. A 120-foot dragon is about 15.24 times the dimensions of the otter; an otter that size would weigh just under 125 short tons. Using the density of a (very light) bird as against a mammal (602 kilos per cubic meter vs. 1,080, i.e. 55.7% the density), that results in a body-weight of a mere 69.6 tons. I'm too tired to compute the
| -1
|
Unrelated
| false
|
<urn:uuid:3b379f95-aadd-4fe3-9d06-50ae3967fc4d>
|
dclm-dedup-25B-ai-scifi-docs | https://www.towerhobbies.com/easyrc/sailplanes/index.html
|
The first step is finding out what you're getting into. Flying RC sailplanes can be quite a commitment in both time and money. Keep this in mind as you go through these pages to make sure you get the most out of this rewarding hobby. A lot of information can be found in this guide but it does not cover everything. If you have a specific type of flying or sailplane style you want to try, look at what is required to fly it. Several beginner sailplanes are suggested at the end of this guide. If you don't want to start with the "real thing," check out one of the best R/C flying simulators available for your PC, RealFlight-X. In addition to a wide variety of sailplanes, it also offers plenty of heli and multirotor aircraft options for you to fly. This is a fun, safe, and worry-free way to learn how to fly R/C aircraft.
RealFlight 8 R/C Flight Simulator
Start Here.
Find an instructor.
There are many designated R/C aircraft flying clubs throughout the U.S. that are supported or chartered by the Academy of Model Aeronautics (AMA). Many of the clubs offer hands-on training and support for beginners and they're a great way to get started in the hobby. Find an AMA-chartered club near you and get started.
Let Tower Hobbies Help.
Tower Hobbies' Phone Sales Staff and Technical Support Staff give you access to 40+ of R/C modeling experience and information. Just call our toll-free number, 1-800-637-6050. We'll help you select the right sailplane! Want to find out what Tower Hobbies has to offer? Request a free Tower Talk sales flyer.
Pick Your Power.
R/C aircraft are powered by many methods that differ in terms of cost and difficulty. Many planes use 2-stroke and 4-stroke engines that burn a methanol/nitro-methane/oil mixture called "glow fuel." For beginners, it is often easier to start with quiet and clean electric motors.
How much does it cost?
The cost of flying depends on your budget. R/C sailplanes often require fewer accessories and less maintenance, so they are often cheaper to purchase and run than R/C powered aircraft. We suggest starting off with an "RTF" or "Ready to Fly" model. These will include everything you need to get you in the air. Most beginner "RTF" sailplanes Tower Hobbies offers will cost around $100-$200.
How fast does a model go?
Most trainers and beginner RTFs usually cruise at 25-30 mph and can land at speeds as slow as 12-15 mph. However, there are also unmodified, off-the-shelf airplanes that can deliver speeds of up to 200 mph! We wouldn't recommend starting with the latter.
How far can a model fly?
The range for a modern R/C system is about a mile. But to maintain control, you need to have your model close enough to see what it is doing. Even a plane with a 6-foot wingspan looks tiny at half a mile. We recommend flying an airplane within your "LOS" or "Line of Sight".
Where Can I Fly?
There are many flying fields located throughout the U.S. that offer great services and facilities for beginners and experts alike. You can check for local sites close to you here. If you don't have a local flying site, please follow the federal, state, and local laws regarding model aircraft. A helpful website to learn how and where you can fly safely is Know Before You Fly. They also have maps that show where model aircraft flight is restricted.
Thermal Sailplanes
Slope Soarers
Tail Types
Conventional Tail
V-TailStabilizer is bent into an upward V shape, and there is no rudder. A radio with mixing capabilities is usually required.
The Model
Hand Launch
Hand Launch Sailplanes
Two-Meter Sailplanes
Open Class
Open Class Sailplanes
Electric Sailplanes
Electric Sailplanes
How do you know what sailplane to choose?
Choosing A Flying Site
Preparing For Flight
R/C Airplane Radio
Radio Control Functions
Check Your Radio Controls
Check Control Surfaces
Adjust Trim
Hand Launch
Land Your Airplane
Helpful Links
| -1
|
Unrelated
| false
|
<urn:uuid:a5ac6c2c-0e4a-40bf-90ec-6b623bd03f53>
|
dclm-dedup-25B-ai-scifi-docs | http://studiofynn.com/journal/future-data-center-and-elevating-brand-values-oracle
|
The Future of the Data Center and Elevating Brand Values ; for Oracle
Visions of the future. Hal 9000 from Stanley Kubrick's 2001, A Space Odyssey
The project initiated with of a research phase to undersand the evolution of the data center and how product design not only impacts the immediate functional requirements but also how it can be a vehicle to communicate the wider notion of technology and the representation of progress. StudioFYNN then guided a workshop with Oracle's design and human factors teams concluding with a strategic roadmap for future product design programs and intiaitives.
In today's competitve and rapidly progressing business environments it is import to understand how brand values connect by association (advertising and communication) and manifestation (products and services). Understanding this difference is key to implementing brand values successfully across all aspects of a business.
| -1
|
Unrelated
| false
|
96948f31-b9a6-409a-aaa1-f4c86a89ea29
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
On expected utility, part 4: Dutch books, Cox, and Complete Class
(Cross-posted from Hands and Cities)
Previously in sequence: Skyscrapers and madmen; Why it can be OK to predictably lose; VNM, separability, and more
This is the final essay in a four-part series on expected utility maximization (EUM). This part focus on theorems that aim to justify the subjective probability aspect of EUM, namely: Dutch Book theorems; Cox’s Theorem (this one is still a bit of a black box to me); and the Complete Class Theorem (this one also supports EUM more broadly). I also briefly discuss Savage, Jeffrey-Bolker, and a certain very general argument for making consistent trade-offs on the margin – both across goods, and across worlds.
I. Comparing with the urns
So we’ve seen three ways of arguing for EUM – an argument from the vNM axioms, an argument from the general connection between separability and additivity, and Peterson’s “direct argument.” In all of these cases, though, we had to assume some probability assignment. Let’s look at that assumption more directly.
The “hanging out with a coin-flipping, urn-pulling God” set-up made the assumption of a probability assignment relatively innocuous, in virtue of the fact that basically everyone wants to be a standard probabilist about things like coins, urns, and spinning wheels. For other types of propositions, though (e.g., “what’s the chance that some human walks on mars before 2100?”), some people, and some theories of probability (see here), start saying: “no, you can’t put probabilities on things like that.”
Still, fans of EUM often do. Indeed, they start putting probabilities on basically any kind of proposition you want -- probabilities often understood to express some subjective level of confidence, and hence called "subjective probabilities." This section briefly describe a more way of thinking about this practice I often use in my own life (I also gestured at this in part 2). Then I turn to some prominent theorems that fans of subjectivity probability often look to for support.
Sup
| -1
|
Unrelated
| false
|
<urn:uuid:e3659c25-50a9-4908-9e82-aec540497a58>
|
dclm-dedup-25B-ai-scifi-docs | http://store.steampowered.com/news/?appgroupname=Dead+Space+Pack&appids=17470,47780
|
Dead Space™ 2
Dead Space 2 is available now on Steam for just $5.00/£2.49—but be quick; it won't stay that price for long.
Dead Space 2 sees you reprise your role as Systems Engineer Isaac Clarke three years after the horrific events on the USG Ishimura. Expect more of same atmospheric horror and monster-stomping action, only this time around, Clarke's a little less taciturn when he encounters more of those murderous Necromorphs.
If you've never explored the original game, you can add that to your library for cheap, too. The Dead Space bundle—which boasts the tense prequel, too—is available for just £3.74, a whopping 81 percent off its usual retail price.
Last month EA closed Visceral Games, the studio behind Dead Space and Battlefield Hardline. Electronic Arts' vice president Patrick Soderlund confirmed the closure and confirmed that the design direction of Visceral's Star Wars project will undergo a "significant change."
In the words of one plucky commenter on the Dead Space 2 customer reviews page, "Don't be sad that it's over, be happy that it happened."
Dead Space - Valve
Today's Deal: Celebrating the launch of the Dead Space animes, Downfall and Aftermath on Steam, save up to 75% on the Dead Space™ series!*
*Offer ends Tuesday November 28th at 10AM Pacific Time. Video content not available in all territories.
Dead Space
The environments of massive open-world games, particularly in recent years, have been rightly praised for their representation, scale and design accuracy. However, there are some gems at the other end of the spectrum - environments that make you feel cramped, tense and desperate for a break. This is an approach to environment design utilised in our real-world, from gardens to architecture, and is mirrored excellently in some game environments, creating areas that trap us in cramped, claustrophobic conditions.
The underground tunnel network of the Metro series, adapted for human life but traversed with trepidation and tension, nailed its own post-apocalyptic look and feel, and had claustrophobia, discomfort and fear oozing from its design. These spaces successfully evoke real-world design principles of landscape mazes and labyrinths, such as dead ends, twists and turns to cause doubling back and elevate desperation, fluctuating size and scale of spaces, and a continuous and monotonal finish (a symphony of grey in Metro's case) that makes every surface and area look the same, but also makes for an unrelenting and repressive aesthetic.
Often, the spaces are not only characteristic of uncomfortable mazes and tunnels, but their disrepair and crumbling structure means they have a constant feeling of pressure and weight about them: the feeling that, at any moment, the space could collapse on top of Artyom's head. The tunnels are also powerful spaces as they are a believable and familiar environment to us; adapting a real-world, recognisably claustrophobic environment makes for a powerfully uncomfortable virtual space.
Read more…
Dead Space™ 2
Electronic Arts closed the doors on Visceral Games yesterday, bringing to a close the studio whose (relatively) recent work includes Dante's Inferno, Army of Two: The Devil's Cartel, Battlefield Hardline, and the Dead Space series. Following the announcement, Zach Wilson, who worked briefly at the studio as a designer on Hardline, took to Twitter to offer some thoughts about where it all went wrong.
Wilson's opening tweets say it all:
In follow-ups, Wilson said that he's "back of the napkinning" the numbers, but added that they're close to the real thing. "I don't know the exact marketing budget but they're frequently close to the dev cost (I have heard this anecdotally)," he wrote.
Sky-high budgets are also why developers are launching their own digital storefronts, rather than simply relying on the simpler and far more ubiquitous Steam.
"Do you hate uplay? Well, the pub gets 90% of the $$$," he tweeted. "EA makes $30 per copy after retailers and console makers take their cut. Then consider that a chunk of the game was sold on sale ... Through Origin they get 90%"
Wilson's tweets don't directly address the reasons for Visceral's closure, but they do paint a very grim portrait of the state of the business, and the extent to which major publisher releases are either big hits, or big busts. If a mid-tier game like Dead Space 2 can knock out four million copies and still be considered underperforming (and keep in mind that EA reported two years after its release that the original Dead Space had sold roughly half that number), then the bar is incredibly high. Any new project that looks like it won't be a huge hit, or won't have a long tail via microtransactions and DLC, suddenly starts to look like a risk from that perspective.
And if, on top of that, the game in question appears to be in trouble, as Kotaku suggested in its report of Visceral's closure yesterday, then risk-averse publishers (which is to say, all of them) aren't likely to wait too long before they take action.
Dead Space - (Adam Smith)
EA have just announced that they’ll be “ramping down and closing” Visceral, the studio behind the Dead Space trilogy. Visceral have been working on an untitled Star Wars project, described as an “action-adventure”, and Amy Hennig, formerly of Naughty Dog and Crystal Dynamics, moved to the studio in 2014 to work on that project as senior creative director. EA’s statement regarding Visceral’s closure suggests that they’re unhappy with the status of that game and they plan to “pivot the design” to fit “fundamental shifts in the marketplace”. Full statement and thoughts below.
Sep 1, 2017
Dead Space
So you're looking to spook yourself with the best horror games you can play on PC. Whether you're into jump scares, interactive fiction, thematically interesting stories or just large men running after you with a chainsaw, we've included a wide variety of games that'll hopefully freak you the hell out. Enjoy.
Like our lists of best strategy games or best FPS games, we tried to focus on a variety of horror experiences that still hold up well today, though we've expanded the remit slightly to include a few retro curios as well.
Resident Evil 7
What starts as a bold, scary reboot certainly gets closer to the more recent action-oriented entries in its later chapters, but exploring the Baker family's grimy plantation in Resident Evil 7 is a grisly treat. The detail of this setting is amazing, and in the first half of the game, there's such a sense of the unknown that you're cautiously poking around every corner and treating bullets like they're gold. Resi 7's videotapes, which have you play out-of-context asides shedding more light on the Baker family and the story, offer the game's best and most experimental moments.
Resi 7 is close to the original intent of Resi, but we kept the HD version of the original on this list too because they're both fantastic in their own way.
An unrelentingly bleak platformer that puts you through a gauntlet of hellish imagery: creepy mermaids, security robots, people hunting you down, nasty weather and more that we won't spoil here. Its vision of a cruel dystopian world that's out to kill you at all times is extraordinary, even if the moment-to-moment platforming is pretty familiar and can be frustrating. You're mainly playing it to experience the setting, really.
See also Little Nightmares, a similar type of horror platformer that isn't as scary but is arguably just as inventive.
Stories Untold
In this anthology game, you operate a computer within the game: first playing an old horror text adventure game set in a spooky house, and later performing similar interactions in other locations, including a lab and a station in freezing conditions. How these episodes link together is the game's overarching mystery, but it's the way the surrounding environment changes with the story beats that'll shit you up here. Stories Untold is co-developed by Alien: Isolation UI mastermind Jon McKellan, and a lot of that DNA is present here. Plus, it'll only take you a few hours to beat, and it's a very reasonable $10 on Steam.
Outlast 2
As a trial-and-error stealth game, Outlast 2 might not be for everyone, but thematically it's among the more interesting games on this list. Playing as a journalist searching for a missing woman in Arizona, your wife is then kidnapped early on by a deranged cult, the origins of which are told through snippets of letters during the game. You navigate dark environments using the night vision mode of your camera, and it's just scary as heck, with a whole village wanting you dead and some of the most gruelling imagery ever put into a game.
System Shock 2
Before BioShock was BioShock, it was System Shock: an altogether freakier combination of RPG and FPS, and one that in its second (and best) iteration told the story of a rogue AI on a haunted spaceship—that rogue AI being the incomparably uppercase SHODAN. The murderous artificial consciousness paved the way for GlaDOS of course, but its the combination of meaningful character advancement, rewarding exploration, horrifying enemies and (at the time) the novel use of audio diaries that make System Shock 2 such a memorable horror game. It was essentially Deus Ex on a spaceship—if you've ever played Deus Ex, or been on a spaceship, you can imagine how delectable that sounds.
Don't be put off by IMSCARED's rather tedious "A Pixelated Nightmare" tagline—it is easily one of the most unsettling games available today. But it's also a tough one to pitch, because much of its terror lies in the surprises that shouldn't be ruined by a meagre 150 word-long recommendation. Know that it borrows from 90's horror games via its aesthetic and fourth wall-breaking, file-bothering makeup; and that it consistently strives to surprise and keep players guessing. Understand that it'll play with your emotions, and drop you into a confused and confusing world while incessantly goading you till its final breath. Don't expect jump scares, but do expect to be scared enough to jump from your chair. The 2012 GameJolt version of IMSCARED is free, while the full, extended version is cheap as chips over on Steam. If you think we're at all grandstanding here, please be our guest and give it a try. We'll be hiding behind the couch.
A rhythm action nightmare in which you play a silver beetle speeding down a track into the mouth of a huge demented boss head. Death comes quickly. Miss a couple of turns and you're dashed into a million glittering pieces against the courses metal banks. Miss a beat in the gaze of the ring-shaped guard robots and they'll hurtle towards you, lasers blazing. All the while the ambient soundtrack pulses uneasily and the the rhythms become faster, and more erratic. The effect is one of tense, compressed dread. Probably best to play it in short bursts only.
Silent Hill 2
We can all agree that Silent Hill 2 is the best in the series, and although Konami have never made much of an effort with the PC versions, if you factor in mods and texture/resolution tweaks this is probably the best way to play it these days—even if prices for the (extremely rare) retail copies can be pretty extortionate. It was the first game to really push the idea of horror narratives as subjective, fluid and untrustworthy things, with a story that invites interpretation and a semi-sentient city that warps and shifts itself to fit the damaged psyches of its inhabitants. The confusing cult nonsense of the first and third games was pushed to the backburner for the more personal story of a psychologically damaged widower battling his way through a foggy purgatory populated by zombie-things, dog-things, and whatever the hell Pyramid Head was.
Whereas the likes of Silent Hill and Fatal Frame rely on radios to alert players to otherworldly adversaries, Sylvio uses sound, EVP (electronic voice phenomenon) and audio manipulation as its central ideas. Not only that, the game builds its entire gorgeously creepy world around this principle theme as players strive to uncover its backstories, bizarre plot twists, and insights into its unsettling unknown—all of which is backed up by some stellar voice acting. Generic first-person horror this ain't, and while it does occasionally force tedious combat set pieces upon players, it thrives in its quirky, idiosyncratic moments that are filled with atmosphere and character and dread. Sylvio is a thinking game and is unique within the horror genre.
Horror games owe a significant debt to one Howard Phillips Lovecraft, and not just because he's long dead and his work is out of copyright. Plenty of games have included references to his unique brand of cosmic horror, but Anchorhead is more inspired than most, drawing from several of his novels and stories to tell the tale of the a married couple who have inherited an old mansion in a creepy New England town. The sedate exploration of the game's opening segments eventually give way to tense, turn-limited puzzles as you struggle to stop an ancient, possibly world-ending ritual from being completed. No pressure then. It's free, and you can play it in your browser.
Amnesia: The Dark Descent
The Dark Descent casts you as Daniel, an amnesiac who wakes up in a mostly deserted castle that must be explored in search of escape. Frictional draw on all of their experience creating atmospheric, exploratory horror in the Penumbra series to fill Amnesia's fortress with an oppressive and lingering sense of foreboding. Expect distant echoing noises, strange rumbles behind the walls, and to start seeing half-formed dark figures in the ambiguous candlelight. There's a monster, too, stalking you through the corridors. The perennial rule of horror creatures—that they're less scary once you've seen and understood them—certainly applies here, but Dark Descent is still a must-play horror game.
Dark Souls
You won't find scripted jump scares here. Dark Souls is a lonely, gruelling struggle through a world on the verge of being extinguished. Lordran is a sad and horrifying place to be. You catch glimpses of the gods' old glory, but mostly you're confronting the aftermath of their terrible mistakes, whether it's the nightmare of the Bed of Chaos or the gross parasite eggs of Demon Ruins. The PC port is poor, but most of its visual shortcomings have been solved by the modding community. Start with the DS Fix and pick and choose from the Dark Souls Nexus to get the game into shape.
Dead Space
Dead Space's lanky alien monsters are noteworthy not just for their ability to fit into tiny closets and jump out at passing protagonists, but for the satisfying fragility of their narrow, bony limbs. Dead Space's high concept, back in the first game, was that you're a simple engineer tending to a broken ship, rather than a meaty space marine with miniguns coming out of his chest. Better still, the cutting and cleaving tools your engineer is so practiced with ended up being more rewarding than the traditional machine guns and shotguns of your typical FPS. Worryingly, foes react differently when you snip off certain limbs—a headshot may only make them madder. Oh, there's a batty plot about an alien obelisk that sends people insane, a space cult, and other nonsense. Don't worry about that too much, the room-to-room stalking is super-tense in spite of the flimsy story. Dead Space classic piece of linear horror design that still holds up.
Stalker: Call of Pripyat
Poor Pripyat just can't catch a break. In real life it's been abandoned since the Chernobyl nuclear disaster. In Stalker, it also suffers the indignity of corrupted anomalies and invisible monsters. The entire series has focused on a harsh and desperate struggle for survival. You may be seeking valuable anomalies and treasure, but first you'll need to secure the basics: food, bandages, and weapons. Occasionally you'll enjoy the companionship of fellow travellers around a campfire, but for the most part your exploration of the open world will feel oppressive and lonely. Call of Pripyat is the best and most technically competent game in the series, but the original Shadow of Chernobyl is also worth a look. Don't miss the Stalker: Lost Alpha—Director's Cut, either.
The Walking Dead
Is anyone still scared of zombies? Sure, they're creepy—there's something intrinsically unsettling about a vacant sack of human flesh—but when is the last time you felt visceral, gut-wrenching fear in the presence of the horde? Blood, guts, and realistic subsurface glistening just don't do it any more. Telltale's The Walking Dead forgoes the anatomy lesson for something more harrowing. The eponymous dead are extras in a bleak human drama, a handy plot device to prompt the fall of society and watch what happens when people break. Those people, all well-written and interesting characters, make for a more immediate, more believable horror story. The Walking Dead could be real, a plausible portrayal of a world going to hell, and that is scary indeed.
Phantasmagoria is the most infamous horror adventure of the interactive movie age, but that's only because almost nobody played the infinitely gorier, endlessly more disturbing Harvester. You wake up with amnesia in a messed up 50s town, where mothers pop their babies' eyeballs, the paperboy packs a gun, the local teachers deals discipline with a baseball bat at Gein Memorial High School, and nobody bats an eye at the wasp woman down the street. All you know is that unless you join the mysterious Lodge in the middle of town, you're not going to last the week—one that ends in an involuntary blood drive where the nurse uses a scythe. Then things get really weird. It's a tough game to find legitimately, but check out our feature on it for more.
Pathologic is ugly and broken. It will sit on your hard-drive like a gangrenous limb, in need of amputation. If this sounds like a criticism, it isn't. Beyond the dirty, putrefied atmosphere, Pathologic is also weird and theatrical, frequently breaking the fourth wall and questioning your role as the player. You choose one of three characters, each with their own mysterious past. Afterwards, masked figures explain the rules of the game: that you have twelve days to cure the town of its disease, and that time will progress regardless of your actions. As it slips by, you'll have to pick your goals wisely, gathering resources and helping characters in the hope of slowing the inexorable decay. Whatever your choice, the town continues to rot, and the game builds towards its horrific conclusion.
It's being remade and expanded, in Pathologic 2, but you can also grab the HD edition of the original on Steam.
Condemned: Criminal Origins
The Silent Hill series does creepy mannequins well, but nowhere near as well as Condemned: Criminal Origins. The premise is quite simple: there's a serial killer on the loose, you're a crime scene investigator, and people expect you to catch him. What's less straightforward is how quickly agent Ethan Thomas takes to cold-blooded murder—even considering the entire populace of Metro City appears to have it in for him. Nonetheless, while Condemned: Criminal Origins offers frontman Thomas a range of firearms, he seems happy enough to do his crowd controlling by way of melee weaponry, each of which has its own distinct feel in close-quarters combat. With that, Condemned rarely pulls any punches—it knows what it is and is happy doing so from start to finish. It's now somehow ten years old, but it holds up well today.
The Evil Within
Reasons to be interested in this survival horror can be boiled down to just two words: Shinji Mikami, the designer responsible for Resident Evil (the good ones), God Hand and Vanquish, the latter of which have criminally never punched and rocket-boosted their way to PC. The Evil Within is his grand return to horror. Expect to spend a fair bit of time hiding from chainsaw-wielding psychopaths, shooting and burning lumbering zombie-likes and laying traps. And a follow-up is on the way, with multiple routes through levels and a story that's a little Silent Hill 2-esque.
Frictional Games has already appeared in this roundup, and that's because time and time again they've proven that they know horror—first in the Penumbra games, and then again in Amnesia. Soma is their latest first-person scare-'em-up, full of creepy experiments, creepier blinking computer-things, and exchanges that question the nature of humanity and consciousness. There are disturbing monsters too, though you can switch those off with a Steam Workshop mod if you want to.
Metro 2033
Similar to Stalker featured earlier in this list, Metro 2033 visits a post-apocalyptic, nuclear war-ravaged world that's filled with mutated abominations—the vast majority of which seek to harm you. Here, the year is 2033, 20 years after Russia fell victim to nuclear war. Moscow's surface is now too dangerous to explore, therefore much of the game takes place within its interweaving subway system and a hostile group named the Dark Ones stalks the player and their pals. Admittedly, Metro 2033, like Stalker, leans towards the action genre however while much of its scare factor is tied to running out of supplies and/or ammo, there's something truly unsettling about its post-nuclear war premise—that perhaps because this sort of scenario could happen, it becomes scarier? Maybe it's simply the fact the Dark Ones are bloody terrifying.
The slim, suited menace known as Slenderman started life as a forum meme, and has quickly grown into a horror series. His schtick is simple, but terrifying enough. If you look directly at him, he devours you, but when you look away he can move position instantly in an attempt to trick your gaze. You have to collect eight notes from a dark forest as the demon hunts you. The free downloadable version, The Eight Pages, has inspired a wealth of YouTube Let's Play videos, because it turns out it's almost as fun to watch Slender's potent psychological terror inflicted on others as it is to endure it yourself. Its popularity encouraged Blue Isle studios and Parsec productions to create a prettier version called Slender: The Arrival, which is available for $10 on Steam, and has bonus Oculus Rift support for VR terror.
Alien: Isolation
The best Alien game ever, by a long way, Isolation stars the smartest, scariest enemy in any game. The Xenomorph's killer instinct is matched only by its curiosity. It learns more about the Sevastopol's nooks and crannies as it hunts you over the course of 12 hours, ripping doors off closets and peering under tables in search of prey. The motion tracker can help you to avoid its grasp, but it can sense the sound, and even the gentle green light of its screen, making every glance a risk. When the game forces you into the vents and you can hear the creature in there with you, Isolation becomes one of the scariest games ever made.
Call of Cthulhu: Dark Corners of the Earth
Lovecraft's Cthulhu mythos should be a ripe playground for gaming scares. It rarely works out like that; the fiction often put to use in ways that fail to convey the sheer magnitude of its ancient and maddening horror. Despite the bugs and the clunkiness, Call of Cthulhu: Dark Corners of the Earth is a first-person survival horror that both stays true to its source, and provides a multitude of ideas through its many and varied levels. You'll go from escaping an assassination, to being hunted by cultists, to fighting off Shoggoths and Deep Ones.
FEAR is a better shooter than a horror game, but is worthy of note for referencing Asian cinema with its creepy villain, Alma, a little girl who can rip people apart with her thoughts. FEAR also exploited the first person perspective to create jump-scares, using ladders and narrow corridors to funnel the player's view through a rollercoaster of linear frights. You catch glimpses of Alma in the corner of a room as lightbulbs shatter, you'll suddenly see her feet at the top of a ladder as you descend, and there's a gratuitous corridor of blood, because The Shining deserves a nod every now and then. First person horror techniques have been honed into a more concentrated horror experience by games like Outlast, but FEAR does let you pin clone soldiers to walls with a stake gun, and kick them in the face in slow motion as they scream “FUUUUUUU” in a low-pitched, slurry expression of terror. The psychological horror themes persisted in FEAR's sequels—FEAR 2: Project Origin and FEAR 3.
Resident Evil HD Remaster
The tank controls and pre-baked backgrounds hint at Resi's age, but it's a survival horror classic nonetheless, and received a handsome HD upgrade in early 2015. The famous Resi mansion drips with atmosphere, and hides some top-drawer jump scares—when crows come crashing through a window, it makes every future trip down that corridor especially tough. The giant spiders are hideous and the relentless threat of the mansion's zombie population grinds down your spirit and your health bar. Soon you're a limping picture of pain and regret, searching for the octagonal object you need to go in the octagonal slot. What a nightmare.
Dead Space - (Adam Smith)
Nostalgia is supposed to be about the things of our early years, but recently I’ve been feeling nostalgic about games released much more recently than my usual rose-tinted diet of Ultima and Daggerfall. I’ve only gone and started missing Dead Space like it was a childhood friend. … [visit site to read more]
Shacknews - Ozzie Mejia
With 2013's Injustice: Gods Among Us, NetherRealm Studios showed it was more than capable of putting together a solid fighting game featuring the DC Comics pantheon. More than that, the developer capably made it stand out from sister franchise Mortal Kombat with its own distinct features and mechanics. For an encore, Injustice 2 further builds on the foundation set by its predecessor and steps forward as a truly superheroic effort.
For Justice
The original Injustice featured a grand DC Elseworlds narrative of a world ruled by a totalitarian Superman and while that plot featured several twists and turns, Injustice 2's Story Mode seamlessly continues this tale with an easy-to-follow recap. Injustice 2 takes place in a post-Regime world with Superman safely imprisoned and Batman desperately trying to pick up the pieces of a world still shaken to its core. His efforts are quickly stalled by a supervillain gathering, the Injustice world's take on the classic Society, which would herald the arrival of extraterrestrial threat and perennial Superman baddie, Brainiac.
With Batman's team overwhelmed, there's a certain pattern the narrative starts to follow. The heroes have had their differences and it looks like now it's time to come together to take on the real villains. Then everyone hugs it out and everything's okay again? In actuality, much like the rest of NetherRealm's Injustice narrative, things aren't that simple. The story that unfolds surrounds the increasing complexity of the classic good vs. evil conflict, as well as what it means to truly deliver justice. It's an eye into Batman's perfectionist (and somewhat naive) view of what justice should be. It's a continuing look at why Superman has pursued the path that he has (albeit one that does Wonder Woman's character a disservice by making her into a Lady MacBeth type). But more than anything, it's the desire from all sides for things to be the way they were and the heartbreaking realization that there is no going back. Fixing things isn't as easy as remembering that everyone's mother happens to be named Martha.
Of course, between all the deeper themes, there's an outstanding, action-filled superhero story that culminates in epic battles unfolding through cutscenes and through standard gameplay.
One big improvement from the first Injustice game is that the quick-time events of the original story are gone. Instead, they're replaced with chapters that center around two characters. Whenever a fight is cued up, the player selects between one of the two heroes, with some of the story's dialogue unfolding differently depending on the character chosen. The choices take a much more extreme turn towards the end of the game, but the story remains cohesive throughout.
Crisis on Infinite Earths
Besides the Story Mode, Injustice 2 also offers the standard single-player mode, but this one comes with a bit of a twist. Playing off the Brother Eye satellite used in the game's story, Multiverse mode offers up contains the standard Arcade mode, where players take on one opponent at a time. However, there are also other Earths that open up with different scenarios and different opponents. Beyond having their own versions of the game's fighters with their own distinct looks, Multiverse mixes things up by occasionally tossing in game-altering conditions, like hazards, boosts, or souped-up opponents.
Multiverse is a great example of using an online connection for something positive, adding in new worlds every day with a finite time to complete their missions and collect their rewards. Some of those worlds have rarer rewards that are worth pursuing. The game even offers a social element to help make Multiverse hopping a little more fun with the Guild system, where groups of friends can earn rewards by completing specific Multiverse tasks. The whole Multiverse package is a robust expansion of the Arcade Mode concept that gives it a much longer life, though anyone with an offline connection can still play the normal Battle Simulator.
Clothes Make the Hero
Let's discuss those Multiverse rewards. They come in the form of Mother Boxes, as Injustice 2 is the latest game to get into the mystery loot craze. With a full comic book universe to play with, the Mother Box rewards dig into the rich DC Comics lore and give each character a dapper new look. The most interesting element, however, is that each piece of gear offers RPG-style boosts. Some of them offer advantages specifically for Multiverse Mode, while others can help give a little boost during multiplayer. What makes the gear system particularly cool is that it gives the sense that a player's fighter is progressing and growing over the course of invested hours. There's a sense of something to aim for, in addition to the usual character ending and the like.
The one problem with the gear system is that oftentimes, a cool Epic or Legendary item will get opened up, but won't be accessible until that particular figher levels up. And sadly, the characters level up about as slowly as the Batmobile with a flat tire. Getting to actually play with that awesome piece of gear will mean hours of grinding and that's when the Multiverse can start to feel tedious.
There's a sense of accomplishment once that gear is finally available, especially for those that want to take it online or assign it to an AI squad that can fight other players' AI squads in simulated combat. The latter is a particularly nifty feature that adds a fantasy element and a cheap way to earn extra experience or loot.
Of course, those that are looking for a more even playing field can also play multiplayer without gear benefits. After all, some people just want to see who's the better player without fancy toys.
Fists of Fury
Speaking of Injustice 2's fighting system, it feels like a further evolution of NetherRealm's distinct fighting style. A friendly tutorial explains everything in great detail, gently explaining how to perform combos, overheads, throws, specials, and anything else required of a NetherRealm fighter. Dashes and slides can cover much more ground, with environmental cues also helping cover a full screen's worth of ground in a moment, helping quell the rise of trigger-happy projectile spammers.
Given that combos are often the bane of the novice's existence, Injustice 2's eloquent explanation of the overhead and other moves that bounce opponents off walls is a godsend. The game encourages players to experiment with juggles, whether it's a simple light attack combo or a cool special move that catches an opponent in mid-air. This complements the rest of the game's mechanics smoothly, including the theatrical Super Moves and the returning Clash system. The latter goes a long way towards giving the game its identity, even if they start to feel old upon repeated viewings.
For All Seasons
Injustice 2 feels like a step up from Gods Among Us in every way imaginable. It's a far deeper experience, follows up wonderfully on the last game's story, and also gives reasons to keep coming back for repeat visits. The rich cast of fighters all have their ow
| -1
|
Unrelated
| false
|
14b13ccf-c6c3-4576-91ec-c26f994ac996
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
Request for Steelman: Non-correspondence concepts of truth
A couple of days ago, Buybuydandavis wrote the following on Less Wrong:
> I'm increasingly of the opinion that truth as correspondence to reality is a minority orientation.
I've spent a lot of energy over the last couple of days trying to come to terms with the implications of this sentence. While it certainly corresponds with my own observations about many people, the thought that most humans simply reject correspondence to reality as the criterion for truth seems almost too outrageous to take seriously. If upon further reflection I end up truly believing this, it seems that it would be impossible for me to have a discussion about the nature of reality with the great majority of the human race. In other words, if I truly believed this, I would label most people as being too stupid to have a real discussion with.
However, this reaction seems like an instance of a failure mode described by Megan McArdle:
> I’m always fascinated by the number of people who proudly build columns, tweets, blog posts or Facebook posts around the same core statement: “I don’t understand how anyone could (oppose legal abortion/support a carbon tax/sympathize with the Palestinians over the Israelis/want to privatize Social Security/insert your pet issue here)." It’s such an interesting statement, because it has three layers of meaning.
>
> The first layer is the literal meaning of the words: I lack the knowledge and understanding to figure this out. But the second, intended meaning is the opposite: I am such a superior moral being that I cannot even imagine the cognitive errors or moral turpitude that could lead someone to such obviously wrong conclusions. And yet, the third, true meaning is actually more like the first: I lack the empathy, moral imagination or analytical skills to attempt even a basic understanding of the people who disagree with me
>
> In short, “I’m stupid.” Something that few people would ever post so starkly on their Facebook feeds.
With this background, i
| -1
|
Unrelated
| false
|
<urn:uuid:f8b1aed3-0a40-4691-843c-35a1aad9ee95>
|
dclm-dedup-25B-ai-scifi-docs | http://stackoverflow.com/questions/12138433/object-file-to-binary-code
|
Join them; it only takes a minute:
Sign up
Join the Stack Overflow community to:
1. Ask programming questions
2. Answer and help your peers
3. Get recognized for your expertise
Let's suppose I have a C file with no external dependency, and only const data section. I would like to compile this file, and then get a binary blob I can load in another program, where the function would be used through a function pointer.
Let's take an example, here is a fictionnal binary module, f1.c
static const unsigned char mylut[256] = {
[0 ... 127] = 0,
[128 ... 255] = 1,
void f1(unsigned char * src, unsigned char * dst, int len)
while(len) {
*dst++ = mylut[*src++];
I would like to compile it to f1.o, then f1.bin, and use it like this in prog.c
int somefunc() {
unsigned char * codedata;
f1_type_ptr f1_ptr;
/* open f1.bin, and read it into codedata */
/* set function pointer to beginning of loaded data */
f1_ptr =(f1_type_ptr)codedata;
/* call !*/
f1_ptr(src, dst, len);
I suppose going from f1.c to f1.o involves -fPIC to get position independance. What are the flags or linker script that I can use to go from f1.o to f1.bin ?
Clarification :
I know about dynamic linking. dynamic linking is not possible in this case. The linking step has to be cast func pointer to loaded data, if it is possible.
Please assume there is no OS support. If I could, I would for example write f1 in assembly with PC related adressing.
share|improve this question
Do you know that you can use shared object files? You compile your .c file to a .so, then you load it dlopen() into you program, and get the function pointer dlsym() to the function. Then you can call it. – Didier Trosset Aug 27 '12 at 8:26
Let's forget libc and dynamic linking – shodanex Aug 27 '12 at 8:33
You want the f1.bin thingie loaded dynamically (i.e. in runtime)? Then you have to build a shared library, and use ldopen()+ldsym() or other module loader (like gmodule). Trying to do it some other way is likely to be hard and refused because of potential security threats (executing data segment and so on). – Michał Górny Aug 27 '12 at 8:36
First of all, as other said you should consider using a DLL or SO.
That said, if you really want to do this, you need to replace the linker script. Something like this (not very well tested, but I think it works):
_dummy_start = 0;
.all : {
_all = .;
LONG(f1 - _all);
*( .text .text.* .data .data.* .rodata .rodata.* )
Then compile with:
$ gcc -c -fPIC test.c
Link with:
$ ld -T script.ld test.o -o test.elf
And extract the binary blob with:
$ objcopy -j .all -O binary test.elf test.bin
Probably some explanation of the script is welcome:
• ENTRY(_dummy_start) That just avoids the warning about the program not having an entry point.
• _dummy_start = 0; That defines the symbol used in the previous line. The value is not used.
• _GLOBAL_OFFSET_TABLE_ = 0; That prevents another linker error. I don't think you really need this symbol, so it can be defined as 0.
• .all That's the name of the section that will collect all the bytes of your blob. In this sample it will be all the .text, .data and .rodata sections together. You may need some more if you have complicated functions, in this case objdump -x test.o is your friend.
• LONG(f1 - _all) Not really needed, but you want to know the offset of your function into the blob, don't you? You cannot assume that it will be at offset 0. With this line the very first 4 bytes in the blob will be the offset of the symbol f1 (your function). Change LONG with QUAD if using 64-bit pointers.
UPDATE: And now a quick'n'dirty test (it works!):
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
typedef void (*f1_t)(char *a, char *b, int len);
f1_t f1;
int main()
char *blob = (char*)valloc(4096);
FILE *f = fopen("test.bin", "rb");
fread(blob, 1, 4096, f);
unsigned offs = *(unsigned*)blob;
f1 = (f1_t)(blob + offs);
mprotect(blob, 4096, PROT_READ | PROT_WRITE | PROT_EXEC);
char txt[] = "¡hello world!";
char txt2[sizeof(txt)] = "";
f1(txt, txt2, sizeof(txt) - 1);
printf("%s\n%s\n", txt, txt2);
return 0;
share|improve this answer
_GLOBAL_OFFSET_TABLE_ is probably needed if reference to other symbols (such as a standard library) are made. – AProgrammer Aug 27 '12 at 9:53
@AProgrammer: But the OP specifically says no external dependency, so it is probably not needed. If he does access to any library then he will have to statically link all the libraries in the blob or do the dynamic linking himself... and that would be... complicated. – rodrigo Aug 27 '12 at 9:58
You should consider building a shared library (.dll for windows, or .so for linux).
Build the lib like this :
gcc -c -fPIC test.c
gcc -shared test.o -o
If you want to load the library dynamically from your code, have a look at the functions dlopen(3) and dlsym(3).
Or if you want to link the library at the compile time, build the program with
gcc -c main.c
gcc main.o -o <binary name> -ltest
I'm really not sure about what I will say here, but this could give you a clue to progress in your research ...
If you don't want to use dlopen and dlsym, you can try to read the symbol table from the .o file in order to find the function address, and then, mmap the object file in memory with the read and execute rights. Then you should be able to execute the loaded code at the address you found. But be carefull with the other dependencies you could meet in this code.
You can check man page elf(5)
share|improve this answer
I precisely don't want to use dynamic linking, edited the question accordingly – shodanex Aug 27 '12 at 8:31
dlopen and dlsym don't imply dynamic linking (since your program won't be linked to the library, your binary won't depend on it, and the library won't be necessary during compilation). The function dlopen lets you load a library, and dlsym will return the function address, based on the symbol name you provided. – phsym Aug 27 '12 at 8:36
dlsym means calling the OS provided dynamic linker to analyse the library file, perform mapping etc.... – shodanex Aug 27 '12 at 8:49
I edited my answer with some possible clues – phsym Aug 27 '12 at 8:52
Your Answer
| -1
|
Unrelated
| false
|
aa5228ba-31f5-4cca-8778-5dc9f503057d
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
The Trouble with Bright Girls [link]
The Trouble with Bright Girls (article @ the Huffington Post)
Excerpt:
> My graduate advisor, psychologist Carol Dweck (author of "Mindset") conducted a series of studies in the 1980s, looking at how Bright Girls and boys in the fifth grade handled new, difficult and confusing material.
>
> She found that Bright Girls, when given something to learn that was particularly foreign or complex, were quick to give up; the higher the girls' IQ, the more likely they were to throw in the towel. In fact, the straight-A girls showed the most helpless responses. Bright boys, on the other hand, saw the difficult material as a challenge, and found it energizing. They were more likely to redouble their efforts rather than give up.
The topic of this article seems to relate to several common Less Wrong issues: the nature of human intelligence, and the gender imbalance among LW readers.
I'm not sure how much credence I give to the proposed explanation of the difference in mindsets. It may well have to do with socialization and feedback, but the specific description of feedback that is presented seems a bit too much of a "just-so story" to me. The difference itself is fascinating, though, and I hope more is done to further our understanding of it.
| -1
|
Unrelated
| false
|
<urn:uuid:bf3efd84-6715-40b0-9522-9d5eef472f5c>
|
dclm-dedup-25B-ai-scifi-docs | https://www.thetoptens.com/reasons-hate-my-little-pony-friendship-is-magic/page3.asp
|
Reasons to Hate My Little Pony: Friendship Is Magic
The Contenders: Page 3
41 It is satanic
The pony characters in the show, since they are part of the equine family, are a symbol of Pride, one of the Seven Deadly Sins. Horse is the animal symbol, and ponies count too.
Like ya' know, the ponies love Satan for no real reason - UntitledMan
42 Queen Chrysalis is the only good character
All characters are bad. Actually not bad, AWFUL.
Queen Chrysalis is also AWFUL, STUPID, AND UGLY!
43 Scootaloo is a ripoff of Buttercup
The actress who does the voice of Scootaloo has the same name as me (Madeleine Peters)
I hope Scootaloo kills herself
44 It ruined your childhood V 1 Comment
45 You're supposed to hate it
Who told you to hate a work of fiction that got popular anyway?
46 Flutter Brutter
What's that?
47 Pegasisters
Does this mean gilda and rainbow dash
Girls who like My Little Pony, GET A LIFE! You Retarded Bunch of Losers!
It is a lame name for them stupid fan girls who obsess over Ponies. More like Pega-Sissies.
I hate mlp but then I'm a girl, so I'm joining the cool dudes. - SnowyAqua
HA! Oh my god you said that. Look, just because you hate MLP doesn't all of the sudden mean you are a boy now and certainly not a cool one. There are girls that hate MLP but they don't say "I'm cool and I'm a boy" - Ultron123
V 12 Comments
48 It is brainwashing
It is brainwashing and it causes unusual behaviors, loss of brain, and turning into a brony to cause Internet problems around the world.
This is not the show's fault. Very sadly, every cartoon out there runs the risk of being abused by obsessive lowlifes - it's the viewer's responsibility to enjoy media without going far enough to cause problems for themselves and others around them. - Entranced98
My sister has been brainwashed by this show. YOU WILL NEVER TAKE ME ALIVE YOU STUPID PONIES!
I had a friend of mine who became a brony and as a result became a huge d! ck to my friends and to myself and as a result have isolated from him. Don't watch this stupid cartoon. - KennyRulz244444
V 4 Comments
49 The fanbase
Ugh. I am from Mario fanbase and never have I seen such a terrible fanbase like My Little Pony.
Both Sonic and Mario Fanbase are really good. But My Little Pony Fanbase is overcrowded with Retarded Lunatics. And they make up stuff that is utterly dumb.
Annoying fanbase which thinks it is manly. Since when is MLP supposed to be manly? What the hell just happened to the world?
I've seen many bad fanbases (FNaF, Sonic (somewhat), Steven Universe, etc) but this one has to be at the very bottom as it is the absolute worst... FNaF is a hair behind. - KennyRulz244444
V 4 Comments
50 Putting your hoof down was a mean-spirited episode
Please don't ever watch that stupid episode, because it teaches you to be mean spirited and get jealous with kind people!
51 Rainbow Dash's personality is a rip off of Naruto's
How is she ripping off sonic? Besides she is more of a rip off buttercup from ppgz the anime because they both love sports and put pressure on one of the characters when they try to do sports and both are tomboys and almost share the same voice and get mad about many things. Although, buttercup was way before the new rainbow dash. Buttercup from the ppgz anime was in 2006 while the new rainbow dash appeared in 2010.
Rainbow's personality is a rip off of:
1. Naruto
2. Sonic
3. Buttercup (with less aggressive)
4. Jake Long (From American Dragon if you don't know)
No it is not! She's a sonic ripoff!
I agree too!
V 2 Comments
52 Too many mean characters
I bet ya that every pony episode has at least one mean, rampaging and disturbing pony with facial expressions overload.
Bronies these days make a lot of dumb assumptions. Get the hell out of the The Top Tens you as-wholes! You get terrible examples from the characters in this messed up show, and bully other people who don't like the show so much.
I hate it when those stupid bronies or pegasisters try to comment back to this list and say they love this messed up show! They deserve many negative values on their stupid comments!
Ikr! Twilight, Pinkie Pie, Rarity, Apple Jack, Fluttershy, and Spike need to take a gun and shoot themselves! I hate them all! They are all mean! I wish they were real so I could stab them 99999999999999999999999999 times!
V 5 Comments
53 Flutter shy has serious problems
You Flutter Shy Fans have to check this list out! There are way too many errors you guys are not aware of for your so called "Sweet and Innocent Flutter Shy" character you brainwashed idiots! I definitely agree with this entire list. Thanks for pointing everything out of this Messed Up Lunatic Flutter Shy.
1. Minute Flutter is frightened of anything that moves like her own shadow. (Dragon Shy)
2. She pretends to be nice, sweet and kind.
3. Flutter breaks down into a Maniac Rage.
I love fluttershy. How dare you guys!
V 11 Comments
54 No male protagonist
Unfortunately, you were told Wrong. This is a show for little Girls, not for Men.
I do Not get this. If this is a Show for Men (As I was told. Clearly, it is NOT.), then WHY is the Protagonist a Stupid Lame Girl?
This sucks. I'm not a male however I think they should make a 'My Little Pony: Gaming is cool' and it has half and half; 3 guys and 3 girls. It could teach you WHAT to do and what NOT to do on the internet. It really would be cool.
Yeah,what about spike? -ApplejackFan
V 8 Comments
55 Racism
"Bridle Gossip". How could such Racism get away in such a cartoon that is supposed to promote Friendship! I am an African and this is the most Insulting My Little Pony episode because them ponies judge her just because she is the Race of the Zebra (How is this not Racism? )
"Hearth's Warming Eve" is the best example of Racism on this cartoon.
They are racist to zebras
Bridle Gossip teaches kids that white people are the best people in sociotey - mayamanga
V 2 Comments
56 Never ending gossiping
They never stop TALKING. It is endless problems Bla Bla Bla this, Bla Bla Bla that. No One Cares about your Problems.
Try watching this show with no Sound. You will know what he means.
In Ancient China, women could only speak when spoken to. In this disgraceful cartoon, they will not stop chatting.
Mlp is so dumb... and all the bronies said that mlp was cooler than kung fu panda! how weird is that! Something is wrong with their brains - SnowyAqua
V 2 Comments
57 Gives you nightmares
Do Princesses Dream Of Magic Sheep?
They also give you cancer - UntitledMan
For Whom The Sweetie Belle Tolls.
This episode is so Messed Up.
58 It is not for kids
It WAS for kids but now its more for like 16+ year olds
When I was little at the weird Royal wedding
,when Rainbow dash did the Sonic Rainboom she took her clothes off!
59 It's popular
Whoever told me to shut up, you are being very rude! Besides, I was only sharing my opinion nicely, until you posted rude comments without thinking. You think being rude is fun and entertaining? It doesn't! So get off the internet if you can't put up with different opinions! I will get off the web if you do the same thing. Capisce?
What's wrong with something popular? Are you some kind of hipster? Hipsters hate things that are mainstream and popular and try to take it out on other people.
I think what he means is that it's too popular.
Shut up mr whats wrong with something popular
60 New characters look weird
PSearch List
Recommended Lists
| -1
|
Unrelated
| false
|
0675d4c7-216f-4af0-8e2d-0d1bb34299be
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
Announcing Duncan Sabien as the keynote speaker for the 2022 LessWrong Community Weekend August 26-29!
We will start taking applications on May 1, with rolling responses until all of the spots are filled. At that time an announcement will be made with all additional details.
For four days at the end of August, aspiring rationalists from all around Europe and beyond will gather at the lovely Lake Wannsee near Berlin to socialize, run workshops, talk, and enjoy our shared forms of nerdiness.
This year our keynote speaker will be the fascinating and incisive Duncan Sabien. He currently works at MIRI and is the former director of curriculum at the Center for Applied Rationality.
Duncan regularly posts valuable new essays on Less Wrong. Some examples of his work are:
* In Defense of Punch Bug,
* Lies, Damn Lies, and Fabricated Options
* r!Animorphs: the Reckoning
* He was the primary person preparing the CFAR Handbook
* and also held a talk at EA Global Creating a Personal Autopilot.
Volunteers: We still can use more help to organize the event. A meetup of this scale takes a lot of work to put together, and If you want to be part of the team turning this into reality, write us an email at lwcw.europe@gmail.com.
Also if someone is interested in creating an activity, or object or something else that will improve the atmosphere or location or create cool new opportunities for interaction, for example like an art installation, or like the (enormously fun) cuddle fort that is made each year in the basement, there is funding to help make it happen. Send us an email to explain your idea if you’d like to make something like that.
What the event is like: While the keynote speaker will be awesome, most of the content is created by the participants. On Friday afternoon we will put up four giant daily planners, and by Saturday morning they are usually filled up with dozens of talks, activities, workshops and events that the attendees have decided to run.
Aside from the content the attendees organize, it’s the serendipitous encounters, shared meals, late night discuss
| -1
|
Unrelated
| false
|
cfb6450d-df24-4153-8ac1-84115bb47021
|
alignment-classifier-documents-unlabeled | StampyAI/alignment-research-dataset/special_docs
|
The Precipice: Existential Risk and the Future of Humanity
[]
[image]
Copyright
Copyright © 2020 Toby Ord,
Jacket design by Amanda Kain
Jacket photograph © NASA/JSC/ASU
Jacket copyright © 2020 by Hachette Book Group, Inc.
Hachette Book Group supports the right to free expression and the value of copyright. The purpose of copyright is to encourage writers and artists to produce the creative works that enrich our culture.
The scanning, uploading, and distribution of this book without permission is a theft of the author’s intellectual property. If you would like permission to use material from the book (other than for review purposes), please contact permissions@hbgusa.com. Thank you for your support of the author’s rights.
Hachette Books
Hachette Book Group
1290 Avenue of the Americas
New York, NY 10104
hachettebooks.com
twitter.com/hachettebooks
Frontispiece illustration © Hilary Paynter, 2020
Toby Ord has asserted his right under the Copyright, Designs and Patents Act, 1988, to be identified as Author of this work
Extract from Pale Blue Dot copyright © 1994 Carl Sagan. Originally published in Pale Blue Dot by Random House. Reprinted with permission from Democritus Properties, LLC. All rights reserved this material cannot be further circulated without written permission of Democritus Properties, LLC.
All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without prior permission in writing from the publishers
Bloomsbury Publishing Plc does not have any control over, or responsibility for, any third-party websites referred to or in this book. All internet addresses given in this book were correct at the time of going to press. The author and publisher regret any inconvenience caused if addresses have changed or sites have ceased to exist, but can accept no responsibility for any such changes
A catalogue record for this book is available from the British Library
ISBN: 978-0-316-48491-6 (hardcover), 978-0-316-48489-3 (ebook)
E3-20200205-JV-NF-ORI
CONTENTS
Cover
Title Page
Copyright
Dedication
List of Figures
List of Tables
PART ONE: THE STAKES
Introduction
1. Standing at the Precipice
How We Got Here
Where We Might Go
The Precipice
2. Existential Risk
Understanding Existential Risk
Looking to the Present
Looking to Our Future
Looking to Our Past
Civilizational Virtues
Cosmic Significance
Uncertainty
Our Neglect of Existential Risks
PART TWO: THE RISKS
3. Natural Risks
Asteroids & Comets
Supervolcanic Eruptions
Stellar Explosions
Other Natural Risks
The Total Natural Risk
4. Anthropogenic Risks
Nuclear Weapons
Climate Change
Environmental Damage
5. Future Risks
Pandemics
Unaligned Artificial Intelligence
Dystopian Scenarios
Other Risks
PART THREE: THE PATH FORWARD
6. The Risk Landscape
Quantifying the Risks
Combining and Comparing Risks
Risk Factors
Which Risks?
7. Safeguarding Humanity
Grand Strategy for Humanity
Risks Without Precedent
International Coordination
Technological Progress
Research on Existential Risk
What You Can Do
8. Our Potential
Duration
Scale
Quality
Choices
Resources
Acknowledgments
Discover More
Appendices
Note on the Author
Note on the Type
Further Reading
Bibliography
Notes
To the hundred billion people before us, who fashioned our civilization;
To the seven billion now alive, whose actions may determine its fate;
To the trillions to come, whose existence lies in the balance.
Explore book giveaways, sneak peeks, deals, and more.
Tap here to learn more.
[Hachette Books logo]
[image]
LIST OF FIGURES
1.1 How we settled the world
1.2 The cradles of civilization
1.3 Striking improvements over the last 200 years
2.1 A classification of existential catastrophes
4.1 The number of stockpiled nuclear warheads over time
4.2 World population from 1700 to 2100
5.1 Measures of progress and interest in AI
5.2 An extended classification of existential catastrophes
6.1 How risks can combine
8.1 A timeline showing the scale of the past and future
D.1 How a 10% and 90% risk may combine
LIST OF TABLES
3.1 Progress in tracking near-Earth asteroids
3.2 The probability per century of a supervolcanic eruption
3.3 The probability per century of a stellar explosion
3.4 Estimates of total natural extinction risk via humanity’s age
3.5 Estimates of total natural extinction risk via related species
3.6 The Big Five extinction events
4.1 Where is the carbon?
6.1 My existential risk estimates
PART ONE
THE STAKES
INTRODUCTION
If all goes well, human history is just beginning. Humanity is about two hundred thousand years old. But the Earth will remain habitable for hundreds of millions more—enough time for millions of future generations; enough to end disease, poverty and injustice forever; enough to create heights of flourishing unimaginable today. And if we could learn to reach out further into the cosmos, we could have more time yet: trillions of years, to explore billions of worlds. Such a lifespan places present-day humanity in its earliest infancy. A vast and extraordinary adulthood awaits.
Our view of this potential is easily obscured. The latest scandal draws our outrage; the latest tragedy, our sympathy. Time and space shrink. We forget the scale of the story in which we take part. But there are moments when we remember—when our vision shifts, and our priorities realign. We see a species precariously close to self-destruction, with a future of immense promise hanging in the balance. And which way that balance tips becomes our most urgent public concern.
This book argues that safeguarding humanity’s future is the defining challenge of our time. For we stand at a crucial moment in the history of our species. Fueled by technological progress, our power has grown so great that for the first time in humanity’s long history, we have the capacity to destroy ourselves—severing our entire future and everything we could become.
Yet humanity’s wisdom has grown only falteringly, if at all, and lags dangerously behind. Humanity lacks the maturity, coordination and foresight necessary to avoid making mistakes from which we could never recover. As the gap between our power and our wisdom grows, our future is subject to an ever-increasing level of risk. This situation is unsustainable. So over the next few centuries, humanity will be tested: it will either act decisively to protect itself and its longterm potential, or, in all likelihood, this will be lost forever.
To survive these challenges and secure our future, we must act now: managing the risks of today, averting those of tomorrow, and becoming the kind of society that will never pose such risks to itself again.
It is only in the last century that humanity’s power to threaten its entire future became apparent. One of the most harrowing episodes has just recently come to light. On Saturday, October 27, 1962, a single officer on a Soviet submarine almost started a nuclear war. His name was Valentin Savitsky. He was captain of the submarine B-59—one of four submarines the Soviet Union had sent to support its military operations in Cuba. Each was armed with a secret weapon: a nuclear torpedo with explosive power comparable to the Hiroshima bomb.
It was the height of the Cuban Missile Crisis. Two weeks earlier, US aerial reconnaissance had produced photographic evidence that the Soviet Union was installing nuclear missiles in Cuba, from which they could strike directly at the mainland United States. In response, the US blockaded the seas around Cuba, drew up plans for an invasion and brought its nuclear forces to the unprecedented alert level of DEFCON 2 (“Next step to nuclear war”).
On that Saturday, one of the blockading US warships detected Savitsky’s submarine and attempted to force it to the surface by dropping low-explosive depth charges as warning shots. The submarine had been hiding deep underwater for days. It was out of radio contact, so the crew did not know whether war had already broken out. Conditions on board were extremely bad. It was built for the Arctic and its ventilator had broken in the tropical water. The heat inside was unbearable, ranging from 113°F near the torpedo tubes to 140°F in the engine room. Carbon dioxide had built up to dangerous concentrations, and crew members had begun to fall unconscious. Depth charges were exploding right next to the hull. One of the crew later recalled: “It felt like you were sitting in a metal barrel, which somebody is constantly blasting with a sledgehammer.”
Increasingly desperate, Captain Savitsky ordered his crew to prepare their secret weapon:
Maybe the war has already started up there, while we are doing somersaults here. We’re going to blast them now! We will die, but we will sink them all—we will not disgrace our Navy\!¹
Firing the nuclear weapon required the agreement of the submarine’s political officer, who held the other half of the firing key. Despite the lack of authorization by Moscow, the political officer gave his consent.
On any of the other three submarines, this would have sufficed to launch their nuclear weapon. But by the purest luck, submarine B-59 carried the commander of the entire flotilla, Captain Vasili Arkhipov, and so required his additional consent. Arkhipov refused to grant it. Instead, he talked Captain Savitsky down from his rage and convinced him to give up: to surface amidst the US warships and await further orders from Moscow.²
We do not know precisely what would have happened if Arkhipov had granted his consent—or had he simply been stationed on any of the other three submarines. Perhaps Savitsky would not have followed through on his command. What is clear is that we came precariously close to a nuclear strike on the blockading fleet—a strike which would most likely have resulted in nuclear retaliation, then escalation to a full-scale nuclear war (the only kind the US had plans for). Years later, Robert McNamara, Secretary of Defense during the crisis, came to the same conclusion:
No one should believe that had U.S. troops been attacked by nuclear warheads, the U.S. would have refrained from responding with nuclear warheads. Where would it have ended? In utter disaster.³
Ever since the advent of nuclear weapons, humans have been making choices with such stakes. Ours is a world of flawed decision-makers, working with strikingly incomplete information, directing technologies which threaten the entire future of the species. We were lucky, that Saturday in 1962, and have so far avoided catastrophe. But our destructive capabilities continue to grow, and we cannot rely on luck forever.
We need to take decisive steps to end this period of escalating risk and safeguard our future. Fortunately, it is in our power to do so. The greatest risks are caused by human action, and they can be addressed by human action. Whether humanity survives this era is thus a choice humanity will make. But it is not an easy one. It all depends on how quickly we can come to understand and accept the fresh responsibilities that come with our unprecedented power.
This is a book about existential risks—risks that threaten the destruction of humanity’s longterm potential. Extinction is the most obvious way humanity’s entire potential could be destroyed, but there are others. If civilization across the globe were to suffer a truly unrecoverable collapse, that too would destroy our longterm potential. And we shall see that there are dystopian possibilities as well: ways we might get locked into a failed world with no way back.
While this set of risks is diverse, it is also exclusive. So I will have to set aside many important risks that fall short of this bar: our topic is not new dark ages for humanity or the natural world (terrible though they would be), but the permanent destruction of humanity’s potential.
Existential risks present new kinds of challenges. They require us to coordinate globally and intergenerationally, in ways that go beyond what we have achieved so far. And they require foresight rather than trial and error. Since they allow no second chances, we need to build institutions to ensure that across our entire future we never once fall victim to such a catastrophe.
To do justice to this topic, we will have to cover a great deal of ground. Understanding the risks requires delving into physics, biology, earth science and computer science; situating this in the larger story of humanity requires history and anthropology; discerning just how much is at stake requires moral philosophy and economics; and finding solutions requires international relations and political science. Doing this properly requires deep engagement with each of these disciplines, not just cherry-picking expert quotes or studies that support one’s preconceptions. This would be an impossible task for any individual, so I am extremely grateful for the extensive advice and scrutiny of dozens of the world’s leading researchers from across these fields.⁴
This book is ambitious in its aims. Through careful analysis of the potential of humanity and the risks we face, it makes the case that we live during the most important era of human history. Major risks to our entire future are a new problem, and our thinking has not caught up. So The Precipice presents a new ethical perspective: a major reorientation in the way we see the world, and our role in it. In doing so, the book aspires to start closing the gap between our wisdom and power, allowing humanity a clear view of what is at stake, so that we will make the choices necessary to safeguard our future.
I have not always been focused on protecting our longterm future, coming to the topic only reluctantly. I am a philosopher, at Oxford University, specializing in ethics. My earlier work was rooted in the more tangible concerns of global health and global poverty—in how we could best help the worst off. When coming to grips with these issues I felt the need to take my work in ethics beyond the ivory tower. I began advising the World Health Organization, World Bank and UK government on the ethics of global health. And finding that my own money could do hundreds of times as much good for those in poverty as it could do for me, I made a lifelong pledge to donate at least a tenth of all I earn to help them.⁵ I founded a society, Giving What We Can, for those who wanted to join me, and was heartened to see thousands of people come together to pledge more than £1 billion over our lifetimes to the most effective charities we know of, working on the most important causes. Together, we’ve already been able to transform the lives of tens of thousands of people.⁶ And because there are many other ways beyond our donations in which we can help fashion a better world, I helped start a wider movement, known as effective altruism, in which people aspire to use evidence and reason to do as much good as possible.
Since there is so much work to be done to fix the needless suffering in our present, I was slow to turn to the future. It was so much less visceral; so much more abstract. Could it really be as urgent a problem as suffering now? As I reflected on the evidence and ideas that would culminate in this book, I came to realize that the risks to humanity’s future are just as real and just as urgent—yet even more neglected. And that the people of the future may be even more powerless to protect themselves from the risks we impose than the dispossessed of our own time.
Addressing these risks has now become the central focus of my work: both researching the challenges we face, and advising groups such as the UK Prime Minister’s Office, the World Economic Forum and DeepMind on how they can best address these challenges. Over time, I’ve seen a growing recognition of these risks, and of the need for concerted action.
To allow this book to reach a diverse readership, I’ve been ruthless in stripping out the jargon, needless technical detail and defensive qualifications typical of academic writing (my own included). Readers hungry for further technical detail or qualifications can delve into the many endnotes and appendices, written with them in mind.⁷
I have tried especially hard to examine the evidence and arguments carefully and even-handedly, making sure to present the key points even if they cut against my narrative. For it is of the utmost importance to get to the truth of these matters—humanity’s attention is scarce and precious, and must not be wasted on flawed narratives or ideas⁸.
Each chapter of The Precipice illuminates the central questions from a different angle. Part One (The Stakes) starts with a bird’s-eye view of our unique moment in history, then examines why it warrants such urgent moral concern. Part Two (The Risks) delves into the science of the risks facing humanity, both from nature and from ourselves, showing that while some have been overstated, there is real risk and it is growing. So Part Three (The Path Forward) develops tools for understanding how these risks compare and combine, and new strategies for addressing them. I close with a vision of our future: of what we could achieve were we to succeed.
This book is not just a familiar story of the perils of climate change or nuclear war. These risks that first awoke us to the possibilities of destroying ourselves are just the beginning. There are emerging risks, such as those arising from biotechnology and advanced artificial intelligence, that may pose much greater risk to humanity in the coming century.
Finally, this is not a pessimistic book. It does not present an inevitable arc of history culminating in our destruction. It is not a morality tale about our technological hubris and resulting fall. Far from it. The central claim is that there are real risks to our future, but that our choices can still make all the difference. I believe we are up to the task: that through our choices we can pull back from the precipice and, in time, create a future of astonishing value—with a richness of which we can barely dream, made possible by innovations we are yet to conceive. Indeed, my deep optimism about humanity’s future is core to my motivation in writing this book. Our potential is vast. We have so much to protect.
1
STANDING AT THE PRECIPICE
It might be a familiar progression, transpiring on many worlds—a planet, newly formed, placidly revolves around its star; life slowly forms; a kaleidoscopic procession of creatures evolves; intelligence emerges which, at least up to a point, confers enormous survival value; and then technology is invented. It dawns on them that there are such things as laws of Nature, that these laws can be revealed by experiment, and that knowledge of these laws can be made both to save and to take lives, both on unprecedented scales. Science, they recognize, grants immense powers. In a flash, they create world-altering contrivances. Some planetary civilizations see their way through, place limits on what may and what must not be done, and safely pass through the time of perils. Others, not so lucky or so prudent, perish.
—Carl Sagan¹
We live at a time uniquely important to humanity’s future. To see why, we need to take a step back and view the human story as a whole: how we got to this point and where we might be going next.
Our main focus will be humanity’s ever-increasing power—power to improve our condition and power to inflict harm. We shall see how the major transitions in human history have enhanced our power, and enabled us to make extraordinary progress. If we can avoid catastrophe we can cautiously expect this progress to continue: the future of a responsible humanity is extraordinarily bright. But this increasing power has also brought on a new transition, at least as significant as any in our past, the transition to our time of perils.
HOW WE GOT HERE
Very little of humanity’s story has been told; because very little can be told. Our species, Homo sapiens, arose on the savannas of Africa 200,000 years ago.² For an almost unimaginable time we have had great loves and friendships, suffered hardships and griefs, explored, created, and wondered about our place in the universe. Yet when we think of humanity’s great achievements across time, we think almost exclusively of deeds recorded on clay, papyrus or paper—records that extend back only about 5,000 years. We rarely think of the first person to set foot in the strange new world of Australia some 70,000 years ago; of the first to name and study the plants and animals of each place we reached; of the stories, songs and poems of humanity in its youth.³ But these accomplishments were real, and extraordinary.
We know that even before agriculture or civilization, humanity was a fresh force in the world. Using the simple, yet revolutionary, technologies of seafaring, clothing and fire, we traveled further than any mammal before us. We adapted to a wider range of environments, and spread across the globe.⁴
What made humanity exceptional, even at this nascent stage? We were not the biggest, the strongest or the hardiest. What set us apart was not physical, but mental—our intelligence, creativity and language.⁵
Yet even with these unique mental abilities, a single human alone in the wilderness would be nothing exceptional. He or she might be able to survive—intelligence making up for physical prowess—but would hardly dominate. In ecological terms, it is not a human that is remarkable, but humanity.
Each human’s ability to cooperate with the dozens of other people in their band was unique among large animals. It allowed us to form something greater than ourselves. As our language grew in expressiveness and abstraction, we were able to make the most of such groupings: pooling together our knowledge, our ideas and our plans.
[image]
FIGURE 1.1 How we settled the world. The arrows show our current understanding of the land and sea routes taken by our ancestors, and how many years ago they reached each area.⁶
Crucially, we were able to cooperate across time as well as space. If each generation had to learn everything anew, then even a crude iron shovel would have been forever beyond our technological reach. But we learned from our ancestors, added minor innovations of our own, and passed this all down to our children. Instead of dozens of humans in cooperation, we had tens of thousands, cooperating across the generations, preserving and improving ideas through deep time. Little by little, our knowledge and our culture grew.⁷
At several points in the long history of humanity there has been a great transition: a change in human affairs that accelerated our accumulation of power and shaped everything that would follow. I will focus on three.⁸
The first was the Agricultural Revolution.⁹ Around 10,000 years ago the people of the Fertile Crescent, in the Middle East, began planting wild wheat, barley, lentils and peas to supplement their foraging. By preferentially replanting the seeds from the best plants, they harnessed the power of evolution, creating new domesticated varieties with larger seeds and better yields. This worked with animals too, giving humans easier access to meat and hides, along with milk, wool and manure. And the physical power of draft animals to help plow the fields or transport the harvest was the biggest addition to humanity’s power since fire.¹⁰
While the Fertile Crescent is often called “the cradle of civilization,” in truth civilization had many cradles. Entirely independent agricultural revolutions occurred across the world in places where the climate and local species were suitable: in east Asia; sub-Saharan Africa; New Guinea; South, Central and North America; and perhaps elsewhere too.¹¹ The new practices fanned out from each of these cradles, changing the way of life for many from foraging to farming.
This had dramatic effects on the scale of human cooperation. Agriculture reduced the amount of land needed to support each person by a factor of a hundred, allowing large permanent settlements to develop, which began to unite together into states.¹² Where the largest foraging communities involved perhaps hundreds of people, some of the first cities had tens of thousands of inhabitants. At its height, the Sumerian civilization contained around a million people.¹³ And 2,000 years ago, the Han dynasty of China reached sixty million people—about a hundred thousand times as many as were ever united in our forager past, and about ten times the entire global forager population at its peak.¹⁴
As more and more people were able to share their insights and discoveries, there were rapid developments in technology, institutions and culture. And the increasing numbers of people trading with one another made it possible for them to specialize in these areas—to devote a lifetime to governance, trade or the arts—allowing us to develop these ideas much more deeply.
Over the first 6,000 years of agriculture, we achieved world-changing breakthroughs including writing, mathematics, law and the wheel.¹⁵ Of these, writing was especially important for strengthening our ability to cooperate across time and space: increasing the bandwidth between generations, the reliability of the information, and the distance over which ideas could be shared.
[image]
FIGURE 1.2 The cradles of civilization. The places around the world where agriculture was independently developed, marked with how many years ago this occurred.
The next great transition was the Scientific Revolution.¹⁶ Early forms of science had been practiced since ancient times, and the seeds of empiricism can be found in the work of medieval scholars in the Islamic world and Europe.¹⁷ But it was only about 400 years ago that humanity developed the scientific method and saw scientific progress take off.¹⁸ This helped replace a reliance on received authorities with careful observation of the natural world, seeking simple and testable explanations for what we saw. The ability to test and discard bad explanations helped us break free from dogma, and allowed for the first time the systematic creation of knowledge about the workings of nature.
Some of our new-found knowledge could be harnessed to improve the world around us. So the accelerated accumulation of knowledge brought with it an acceleration of technological innovation, giving humanity increasing power over the natural world. The rapid pace allowed people to see transformative effects of these improvements within their own lifetimes. This gave rise to the modern idea of progress. Where the world had previously been dominated by narratives of decline and fall or of a recurring cycle, there was increasing interest in a new narrative: a grand project of working together to build a better future.
Soon, humanity underwent a third great transition: the Industrial Revolution. This was made possible by the discovery of immense reserves of energy in the form of coal and other fossil fuels. These are formed from the compressed remains of organisms that lived in eons past, allowing us access to a portion of the sunlight that shone upon the Earth over millions of years.¹⁹ We had already begun to drive simple machines with the renewable energy from the wind, rivers and forests; fossil fuels allowed access to vastly more energy, and in a much more concentrated and convenient form.
But energy is nothing without a way of converting it to useful work, to achieve our desired changes in the world. The steam engine allowed the stored chemical energy of coal to be turned into mechanical energy.²⁰ This mechanical energy was then used to drive machines that performed massive amounts of labor for us, allowing raw materials to be transformed into finished products much more quickly and cheaply than before. And via the railroad, this wealth could be distributed and traded across long distances.
Productivity and prosperity began to accelerate, and a rapid sequence of innovations ramped up the efficiency, scale and variety of automation, giving rise to the modern era of sustained economic growth.²¹
The effects of these transitions have not always been positive. Life in the centuries following the Agricultural Revolution generally involved more work, reduced nutrition and increased disease.²² Science gave us weapons of destruction that haunt us to this day. And the Industrial Revolution was among the most destabilizing periods in human history. The unequal distribution of gains in prosperity and the exploitative labor practices led to the revolutionary upheavals of the early twentieth century.²³ Inequality between countries increased dramatically (a trend that has only begun to reverse in the last two decades).²⁴ Harnessing the energy stored in fossil fuels has released greenhouse gases, while industry fueled by this energy has endangered species, damaged ecosystems and polluted our environment.
Yet despite these real problems, on average human life today is substantially better than at any previous time. The most striking change may be in breaking free from poverty. Until 200 years ago—the last thousandth of our history²⁵—increases in humanity’s power and prosperity came hand in hand with increases in the human population. Income per person stayed almost unchanged: a little above subsistence in times of plenty; a little below in times of need.²⁶ The Industrial Revolution broke this rule, allowing income to grow faster than population and ushering in an unprecedented rise in prosperity that continues to this day.
We often think of economic growth from the perspective of a society that is already affluent, where it is not immediately clear if further growth even improves our lives. But the most remarkable effects of economic growth have been for the poorest people. In today’s world, one out of ten people are so poor that they live on less than two dollars per day—a widely used threshold for “extreme poverty.” That so many have so little is among the greatest problems of our time, and has been a major focus of my life. It is shocking then to look further back and see that prior to the Industrial Revolution 19 out of 20 people lived on less than two dollars a day (even adjusting for inflation and purchasing power). Until the Industrial Revolution, any prosperity was confined to a tiny elite with extreme poverty the norm. But over the last two centuries more and more people have broken free from extreme poverty, and are now doing so more quickly than at any earlier time.²⁷ Two dollars a day is far from prosperity, and these statistics can be of little comfort to those who are still in the grip of poverty, but the trends toward improvement are clear.
And it is not only in terms of material conditions that life has improved. Consider education and health. Universal schooling has produced dramatic improvements in education. Before the Industrial Revolution, just one in ten of the world’s people could read and write; now more than eight in ten can do so.²⁸ For the 10,000 years since the Agricultural Revolution, life expectancy had hovered between 20 and 30 years. It has now more than doubled, to 72 years.²⁹ And like literacy, these gains have been felt across the world. In 1800 the highest life expectancy of any country was a mere 43 years, in Iceland. Now every single country has a life expectancy above 50.³⁰ The industrial period has seen all of humanity become more prosperous, educated and long-lived than ever before. But we should not succumb to complacency in the face of this astonishing progress. That we have achieved so much, and so quickly, should inspire us to address the suffering and injustices that remain.
We have also seen substantial improvements in our moral thinking.³¹ One of the clearest trends is toward the gradual expansion of the moral community, with the recognition of the rights of women, children, the poor, foreigners and ethnic or religious minorities. We have also seen a marked shift away from violence as a morally acceptable part of society.³² And in the last sixty years we have added the environment and the welfare of animals to our standard picture of morality. These social changes did not come naturally with prosperity. They were secured by reformers and activists, motivated by the belief that we can—and must—improve. We still have far to go before we are living up to these new ideals, and our progress can be painfully slow, but looking back even just one or two centuries shows how far we have come.
Of course, there have been many
| -1
|
Unrelated
| false
|
c342f07d-62f3-45da-8f6f-cd958150f38f
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
An issue with MacAskill's Evidentialist's Wager
In The Evidentialist's Wager, MacAskill et al. argue as follows:
Suppose you are uncertain as to whether EDT or CDT is the correct decision theory, and face a Newcomb-like decision. If CDT were correct, your decision would only influence the outcomes of your present decision. If EDT were correct, it would provide evidence not only for the outcome of your present decision, but also for the outcome of many similar decisions by similar agents throughout the universe (either because they are exact copies of you, or they have very similar decision theories/computations to you, etc.). Thus, the stakes for the decision are way higher if EDT is true, and so you should act as if EDT were true (even if you have higher prior credence on CDT).
This argument of course relies on how many of these similar agents actually exist, and how similar they are. They use the term correlated agent to mean some agent similar enough to you so that your decision will acausally provide evidence about theirs. As possible counterexamples to their argument, they point out the existence of different agents:
1. Anti-correlated agents: Agents whose decision theory will drive them to take the decision opposite to yours.
2. Evil Twins: Agents positively correlated to you (with your same decision theory) but with drastically different utility functions (or in the extreme, the exact opposite utility function).[1]
Regarding anti-correlated agents, since our decision theories are actual good heuristics for going about the world and rationally obtaining our goals, it seems more likely for agents that exist (that is, have survived) to be positively correlated rather than anti-correlated.
But regarding Evil Twins, because of the Orthogonality Thesis, we might expect on average that there are as much agents positively correlated to us with ~our same utility function (Good Twins) as agents positively correlated to us with ~opposite utility function (Evil Twins).
That is, the universe selects for agents
| -1
|
Unrelated
| false
|
c015e5c9-cc06-4bf8-b4e5-ae3b483699ce
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
Hanging Out My Speaker's Shingle
I was recently invited to give a talk on heuristics and biases at Jane Street Capital, one of the top proprietary trading firms ("proprietary" = they trade their own money). When I got back home, I realized that (a) I'd successfully managed to work through the trip, and (b) it'd been very pleasant mentally, a nice change of pace. (One of these days I have to blog about what I discovered at Jane Street - it turns out they've got their own rationalist subculture going.)
So I've decided to hang out my shingle as a speaker at financial companies.
You may be thinking: "Perhaps, Eliezer, this is not the best of times."
Well... I do have hopes that, among the firms interested in having me as a speaker, a higher-than-usual percentage will have come out of the crash okay. I checked recently to see if this were the case for Jane Street Capital, and it was.
But more importantly - your competitors are learning the secrets of rationality! Are you?
Or maybe I should frame it as: "Not doing too well this year? Drop the expensive big-name speakers. I can give a fascinating and useful talk and I won't charge you as much."
And just to offer a bit of a carrot - if I can monetize by speaking, I'm much less likely to try charging for access to my future writings. No promises, but something to keep in mind. So do recommend me to your friends as well.
I expect that, as I speak, the marginal value of money to my work will go down; the more I speak, the more my price will go up. If my (future) popular book on rationality becomes a hit, I'll upgrade to big-name fees. And later in my life, if all goes as planned, I'll be just plain not available.
So I'm offering you, my treasured readers, a chance to get me early. I would suggest referencing this page when requesting me as a speaker. Emails will be answered in the order they arrive.
| -1
|
Unrelated
| false
|
5c088e20-6f25-4241-8177-2d913273d3ec
|
alignment-classifier-documents-unlabeled | trentmkelly/LessWrong-43k
|
The Best Software For Every Need
When I first started programming, I didn't use a terminal multiplexer and finding tmux was a sort of revelation. I joked once on discord that "life before tmux was not life". It strikes me there are probably many other programs that I am not aware of that would be useful to know about.
I've occasionally found Luke's The Best Textbooks on Every Subject thread useful, so I thought a similar thread about software may be interesting.
Here are the rules:
1. Post the name of a program for a given need.
2. You must have tried at least 2 other programs designed for the same/similar class of problems.
3. You must briefly name the other programs you have tried and why you think your chosen program is superior to them.
| -1
|
Unrelated
| false
|
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 115
Size of downloaded dataset files:
8.26 MB
Size of the auto-converted Parquet files:
8.26 MB
Number of rows:
1,065