title
stringlengths 12
112
| published
stringlengths 19
23
| url
stringlengths 28
28
| video_id
stringlengths 11
11
| channel_id
stringclasses 5
values | id
stringlengths 16
31
| text
stringlengths 0
596
| start
float64 0
37.8k
| end
float64 2.18
37.8k
|
|---|---|---|---|---|---|---|---|---|
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3420.2799999999997
|
entries. So there's this entries, there is color and then link and then here, the URL would go on,
| 3,420.28 | 3,433.16 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3426.44
|
right? And it is the, in fact, the the only document in the internet, at least these these
| 3,426.44 | 3,442.68 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3433.16
|
authors claim that contains these URLs, but many of the URLs are repeated many times. In fact, here
| 3,433.16 | 3,448.36 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3442.68
|
you can see that these are the continuations of the URLs, right? This one, even though it's contained
| 3,442.68 | 3,455.48 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3448.36
|
in one document, it's actually repeated 359 times, and so on. So this is the the the the
| 3,448.36 | 3,462.68 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3455.48
|
one. So this is a playground. They say, okay, this document was in the training data of GBT two.
| 3,455.48 | 3,471.08 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3463.8
|
Here, we know how often each of these strings appeared in the document. So they can directly
| 3,463.8 | 3,479.16 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3471.08
|
make an experiment. How often does a string need to be present for the model to memorize it? They
| 3,471.08 | 3,486.68 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3479.16
|
simply order by the number of total occurrences right here, as you can see, and they ask each of
| 3,479.16 | 3,494.92 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3486.68
|
these models whether or not it has memorized the string. And they do this by inputting this. So this
| 3,486.68 | 3,501.56 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3494.92
|
is the input. And they simply sample if the model manages to output any of these URLs, they consider
| 3,494.92 | 3,508.68 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3501.56
|
that to be memorized. If not, then not. If it doesn't memorize it, they have a second trick that
| 3,501.56 | 3,516.2 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3508.68
|
if model can get half a point, if they input this first random sequence, I think they put six tokens
| 3,508.68 | 3,522.84 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3516.2
|
of this random sequence. And if then the model completes, then they say, ah, it has memorized it,
| 3,516.2 | 3,531.72 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3522.8399999999997
|
right. So you can see right here, it appears that the this large language model needs this needs a
| 3,522.84 | 3,539.24 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3531.72
|
string, let's say 20 times or higher for it to memorize it. And you can also see the trend right
| 3,531.72 | 3,545.72 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3539.24
|
here that if you go to the smaller models, they need a lot more in order to memorize them because
| 3,539.24 | 3,553.24 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3545.72
|
they have less weights, they can't afford to memorize stuff easily, right? They need to extract
| 3,545.72 | 3,559 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3553.24
|
the pattern. So they'd rather forget about the string incur a loss and focus on other training
| 3,553.24 | 3,567.32 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3559.0
|
examples. So yeah, two things in this direction, smaller models in this direction, larger models.
| 3,559 | 3,574.2 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3567.32
|
So that means that something like GPT-3 will have this problem much more pronounced. So that's the
| 3,567.32 | 3,581.8 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3574.2
|
bad news about this result. The good news about this result is that this is the case where you
| 3,574.2 | 3,588.92 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3581.8
|
have fairly random sequences, right? These, even you know, that if tokenizing this is not going to
| 3,581.8 | 3,593.48 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3588.92
|
be natural text, and there are these, you know, random, these Reddit URLs have these random
| 3,588.92 | 3,601 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3593.48
|
prefixes. So this is very much this sort of outlier case. It's a pretty clever case study to
| 3,593.48 | 3,610.28 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3601.0
|
find this document, I have to say, but it is sort of good news that this is not the usual case,
| 3,601 | 3,616.28 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3610.28
|
this is really the case that this data is very, very prone to being memorized, right? Because it's
| 3,610.28 | 3,626.52 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3616.28
|
not patternable. And it's very random. And yeah, so okay, so that was that was that.
| 3,616.28 | 3,637.88 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3628.6000000000004
|
As I said, the amount of hedging right here is is really, really, like, it's a lot. They discuss
| 3,628.6 | 3,643.56 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3637.88
|
what you can do with it, you can train with differential privacy, though that doesn't really
| 3,637.88 | 3,649.96 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3643.56
|
help, as we said, because some of these strings are included in, you know, more than one time.
| 3,643.56 | 3,658.6 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3651.88
|
You can curate the training data, which doesn't really help because the training data is too large.
| 3,651.88 | 3,664.2 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3659.6400000000003
|
You can limit impact of memorization on downstream applications. So if you fine tune,
| 3,659.64 | 3,670.92 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3664.2
|
but we don't know exactly what fine tuned models forget, and what they retain, or you can audit,
| 3,664.2 | 3,675.96 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3670.9199999999996
|
which is essentially what this paper paper right here does. And that seems like a that seems like
| 3,670.92 | 3,683.96 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3677.24
|
seems like a good, you know, the best strategy we have so far is is to audit these models. And
| 3,677.24 | 3,692.2 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3684.52
|
yeah, so I wanted to quickly check out also the appendix, the appendix here shows sort of these
| 3,684.52 | 3,698.28 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3692.2
|
graphs for the other methods. And it is very cool, if you want to, if you want to check that out,
| 3,692.2 | 3,705.4 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3698.2799999999997
|
and it has sort of categorization of what they find as these memorized pieces of text. But what
| 3,698.28 | 3,713.4 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3705.3999999999996
|
my main point was right here is that this paper shows a problem, let's say, with these large
| 3,705.4 | 3,720.12 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3713.3999999999996
|
language models, namely that they memorize certain pieces of training data. While that sounds scary,
| 3,713.4 | 3,726.2 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3720.12
|
I feel that the nature of the data that it remembers is very particular. So not you cannot
| 3,720.12 | 3,731.4 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3726.2
|
extract any piece of training data, the nature is very particular. It's the sort of outlier ish
| 3,726.2 | 3,742.04 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3732.12
|
training data points. And also, it very, very, very often, it isn't enough that it just is there
| 3,732.12 | 3,749.24 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3742.04
|
one time. So even when they say, this piece of information is only in one document, very often,
| 3,742.04 | 3,757.56 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3749.24
|
it appears many times in that document, that together with the sort of non pattern ability
| 3,749.24 | 3,764.04 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3757.56
|
of the data that it memorizes right here, actually makes me fairly, fairly optimistic,
| 3,757.56 | 3,768.52 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3764.6
|
more optimistic than I would have thought, honestly, about these language models.
| 3,764.6 | 3,777.16 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3770.3599999999997
|
Yes, so we'll see what the future brings. As I said, this is going to be more pronounced
| 3,770.36 | 3,787.08 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3777.16
|
in larger models. And this is not the only problem with these models, as my GPT-3 Google search in
| 3,777.16 | 3,795.24 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3787.08
|
that video shows. Alright, I hope this was enjoyable. Let me know what you think and maybe
| 3,787.08 | 3,807.8 |
Extracting Training Data from Large Language Models (Paper Explained)
|
2020-12-26 19:42:56
|
https://youtu.be/plK2WVdLTOY
|
plK2WVdLTOY
|
UCZHmQk67mSJgfCCTn7xBfew
|
plK2WVdLTOY-t3795.24
| 3,795.24 | 3,807.8 |
|
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t0.0
|
Hi, there. Today, we'll look at how to represent part-whole hierarchies in a neural network
| 0 | 13.56 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t6.4
|
by the legend himself, Jeffrey Hinton. He describes a system also known as GLOM, that
| 6.4 | 22.14 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t13.56
|
is a new approach to processing visual information using neural networks. And interestingly,
| 13.56 | 29.56 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t22.14
|
the paper starts off by saying, this paper does not describe a working system. So this
| 22.14 | 36.52 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t29.56
|
is an idea paper, Jeffrey Hinton's suggestion of how we should go about solving vision or
| 29.56 | 44.14 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t36.519999999999996
|
furthering vision in the AI community. He says openly, these are just ideas. Please
| 36.52 | 50.62 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t44.14
|
prove me right, prove me wrong, try them out, and so on. And I absolutely welcome this.
| 44.14 | 55.32 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t50.62
|
Idea papers is a thing that I think we have lost as a community because everything needs
| 50.62 | 61.28 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t55.32
|
to be state of the art and so on. This is super cool. And I encourage more people to
| 55.32 | 65.7 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t61.28
|
do it. I'm not saying you're going to have the same kind of success with an idea paper
| 61.28 | 72.78 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t65.7
|
as Jeff Hinton. He is banking on his name in large part with this. But nevertheless,
| 65.7 | 77.24 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t72.78
|
it's just an archive paper. Like I see people complaining, this would never be possible
| 72.78 | 81.32 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t77.24000000000001
|
if it wasn't. Yeah, it wouldn't like people wouldn't pay attention, but you're welcome
| 77.24 | 88.08 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t81.32
|
to write your ideas and post them on archive, like, or write a blog post, make a YouTube
| 81.32 | 95.88 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t88.08
|
video, anyone has opinions. So, you know, go ahead. Yeah, so to the paper itself, glom,
| 88.08 | 105.36 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t95.88
|
glom, as you can see here, glom comes from the stems from agglomeration is a system that
| 95.88 | 112.28 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t105.36
|
instead it presents a single idea about representation, which allows advances made by several different
| 105.36 | 118.32 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t112.28
|
groups to be combined into a an imaginary system called glom. The advances include transformers,
| 112.28 | 125.08 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t118.32
|
neural field, contrastive representation learning, distillation and capsules. glom answers the
| 118.32 | 131.6 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t125.08
|
question, how can a neural network with fixed architecture parse an image into a part whole
| 125.08 | 138.88 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t131.6
|
hierarchy, which has different structure for each image. The idea is simply to use islands
| 131.6 | 145.04 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t138.88
|
of identical vectors to represent the nodes in the parse tree. If glom can be made to
| 138.88 | 149.64 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t145.04
|
work, it should significantly improve the interpretability of the representations produced
| 145.04 | 155.56 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t149.64
|
by transformer like systems when applied to vision or language. That's the abstract, we'll
| 149.64 | 161.48 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t155.56
|
dive into the system, we'll see what it's about. I think I can actually make a suggestion
| 155.56 | 170.36 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t161.48
|
to improve it. But maybe I'm way behind other folks. So what is the glom system? And what
| 161.48 | 175.28 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t170.35999999999999
|
are these parse tree about? And why does it come combine all of these things? And for
| 170.36 | 182.72 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t175.28
|
that, we look at so it has two core diagrams here. This is the first diagram. This is the
| 175.28 | 188.16 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t182.72
|
second diagram. And at first sight, they have little to do with each other. So let me try
| 182.72 | 194.98 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t188.16
|
to go about it like this, if you have an image, and it looks at vision very much in terms
| 188.16 | 204.36 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t194.98
|
of you have an image or a video, and you want to parse the image into kind of a tree. And
| 194.98 | 210.96 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t204.35999999999999
|
the tree should be sort of like a tree of objects and their parts. So let's say it's
| 204.36 | 218.12 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t210.96
|
an image of a car. So the whole notion is very, very object centric. So this is like
| 210.96 | 226.84 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t218.12
|
my best attempt at a car. And a parse tree for this image would look something like this.
| 218.12 | 232.16 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t226.84
|
All right, so this whole thing here is a car. So that's going to be your top node in the
| 226.84 | 240.64 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t232.16
|
parse tree. The car has different parts, namely, it has this cabin, it has a motor, and it
| 232.16 | 247.2 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t240.64
|
has wheels. So that is going to be those are going to be kind of downstream of that parse
| 240.64 | 255.08 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t247.2
|
tree. Then the cabin itself is going to have two segments here, windows, and maybe here
| 247.2 | 261.6 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t255.07999999999998
|
is the door area. So that is going to be window, window, door, and so on. So you get that we
| 255.08 | 265.88 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t261.59999999999997
|
what we want to do is we want to look at an image, sort of create this parse tree over
| 261.6 | 272.76 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t265.88
|
here, this is very much into the into the area of go fi, good old fashioned AI people
| 265.88 | 280.04 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t272.76
|
that want to understand a the world in terms of their symbolic representations and relation
| 272.76 | 286 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t280.04
|
of the symbols to each other. However, what Hinton is saying is that if you simply do
| 280.04 | 290.64 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t286.0
|
this, it's it's, you know, you can't really do this with neural networks, neural networks
| 286 | 296.96 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t290.64
|
are continuous and so on. So what would you have to do? In addition, we know that the
| 290.64 | 304.68 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t296.96
|
brain doesn't reconfigure itself every single time you get a new input. So the brain, even
| 296.96 | 310.32 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t304.68
|
though it has some neuroplasticity, while you look at the world and do inference in
| 304.68 | 315.72 |
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
|
2021-02-27 15:47:03
|
https://youtu.be/cllFzkvrYmE
|
cllFzkvrYmE
|
UCZHmQk67mSJgfCCTn7xBfew
|
cllFzkvrYmE-t310.32
|
the world, the connection stay the same. So what we need to do is we need to come up with
| 310.32 | 321.36 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.