| Not checkable |
Yann LeCun | Not checkable |
2021 |
You have to have somewhere representing a sort of infinite number of plausible continuations of multiple frames in a high dimensional…
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think it's integrated.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
That's called a collapse.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
In the first months of life, babies learn how the world works mostly by observation because they can hardly act in the world.
Yann LeCun · Factual · 2021 |
Philosophy, ethics and society |
| Not researched |
Yann LeCun | Factual |
2021 |
Across AI sub-fields there has been a continuous evolution from less hand-crafting to more learning.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It is, but we're not there yet.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So it's easier to learn when your objective is intrinsic.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So that has to be hardwired.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
People have used data augmentation on MNIST digits since the 1990s.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not researched |
Yann LeCun | Factual |
2021 |
Geoffrey Hinton proposed, in the early 1990s with his student Sue Becker, the idea of maximizing mutual information between the outputs of two systems.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
Supervised learning requires many samples for learning anything.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I don't know how to make logic-based reasoning
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It doesn't tell us how to feed ourselves.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
After self-supervised representation learning, recognizing handwritten digits on top of it requires only a few labeled samples.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It's not platonic because eventually I want the stuff to get used, but it's okay if it takes five or 10 years for the community to realize…
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
The difference between human and chimp genomes involved in human intelligence is less than 1% of the genome.
Yann LeCun · Factual · 2021 |
Science and research methods |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I have kind of an orthodox folk speculative hypothesis about consciousness.
Yann LeCun · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Prediction |
2021 |
If the US does not take a radical step toward a fair, inclusive system, it will not be the leading country in the world for much longer.
Sam Altman · Prediction · 2021 |
AI policy and governance |
| Not researched |
Yann LeCun | Factual |
2021 |
The system under discussion worked, but they ended up not using it.
Yann LeCun · Factual · 2021 |
Technology |
| Not researched |
Yann LeCun | Factual |
2021 |
LeCun used to call the two-identical-copies network architecture a Siamese network.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Unresolvable |
Sam Altman | Prediction |
2021 |
By default, automation will concentrate wealth and power in a tiny number of hands.
Sam Altman · Prediction · 2021 |
AI, the economy and work |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I mean, it's a superficial understanding, obviously.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
We’d also need to figure out rules about transferability and borrowing against this equity.
Sam Altman · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
With a differentiable world model such as a neural net, one can do gradient-based model predictive control by backpropagating through it.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Unresolvable |
Yann LeCun | Prediction |
2021 |
Training systems on video prediction with self-supervised techniques will provide a path towards some level of physical common sense in machines.
Yann LeCun · Prediction · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think the big challenge of AI for the next decade,
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
If we had as many world models as there are situations we encounter, then we could do all of them simultaneously.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
Orangutans are solitary animals that do not seek the company of others.
Yann LeCun · Factual · 2021 |
Other |
| Resolved true |
Sam Altman | Factual |
2021 |
America has repeatedly found ways to challenge concentration of wealth and power.
Sam Altman · Factual · 2021 |
Philosophy, ethics and society |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
Your model of the world should have your actions as one of the inputs, and that will drive you to learn causal models of the world
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
This would harm all Americans more than most realize.
Sam Altman · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
and give more people a fair shot at achieving the life they want.
Sam Altman · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Prediction |
2021 |
Altman said he thinks a model similar to the one used by companies can work for the US as well as it does for companies.
Sam Altman · Prediction · 2021 |
Economics and markets |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
And that's the answer to this probably is self supervised running.
Yann LeCun · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Factual |
2021 |
Countries that concentrate wealth in a small number of families do worse over the long term.
Sam Altman · Factual · 2021 |
Philosophy, ethics and society |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
The essence of intelligence is the ability to predict
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
And they are extremely inefficient.
Yann LeCun · Not checkable · 2021 |
|
| Mostly true |
Yann LeCun | Factual |
2021 |
A cat has fewer than 1 billion neurons, while a dog has 2 billion.
Yann LeCun · Factual · 2021 |
Other |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
There's nothing wrong with that.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
And the main issue with learning world models or learning predictive models is that the prediction is never a single thing because the…
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
Babies learn an enormous amount of background knowledge about the world that may be the basis of common sense.
Yann LeCun · Factual · 2021 |
Philosophy, ethics and society |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
That engine is configurable to the situation at hand.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
The first one is, you know, getting machines to learn to represent the world.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think most people don't actually realize what death is, right?
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It doesn't make the process exist.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
the fundamental input to wealth generation isn’t farmland, but money and ideas
Sam Altman · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
Self-supervised learning offers far more training signal than supervised or reinforcement learning.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
Now, that said, this ability to learn world models, I think is the key to the possibility of running machines that can also reason.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
But I don't think they're fundamental, you know, active learning, even to some extent reinforcement learning.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It's, to me, it seems very implausible
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
This is the kind of problem we've evolved to solve.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So it's the more fundamental question to ask.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
I think that every adult US citizen should get an annual share of the US GDP.
Sam Altman · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It's just statistics of a particular kind.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So I think the action part would be necessary for having causal models of the world.
Yann LeCun · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Prediction |
2021 |
American Equity would create a society that Altman believes would work much better than today's.
Sam Altman · Prediction · 2021 |
Philosophy, ethics and society |
| Not researched |
Yann LeCun | Factual |
2021 |
Generally, the model is mostly deterministic and mostly built by hand, though sometimes a few parameters have to be identified through what is called systems identification.
Yann LeCun · Factual · 2021 |
Technology |
| Not researched |
Yann LeCun | Factual |
2021 |
Barlow Twins is a method for maximizing the information content of a vector.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
And it might suggest the idea that consciousness basically is the module that configures this world model of yours.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
Can we get them to spontaneously learn good hierarchical representations of actions?
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So cats and rats are incredibly curious.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
There would be difficult consequences for things like immigration policy that will need a lot of discussion.
Sam Altman · Not checkable · 2021 |
|
| Mostly true |
Yann LeCun | Factual |
2021 |
Since the 1990s, algorithms exist that learn oriented edge detectors fully unsupervised with the equivalent of a few minutes of real time.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
consciousness is not a consequence of the power of our mind,
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
The first vision systems in animals evolved mainly for localization, with very little recognition.
Yann LeCun · Factual · 2021 |
Science and research methods |
| Unresolvable |
Yann LeCun | Prediction |
2021 |
A machine cannot be trained to be intelligent purely from text.
Yann LeCun · Prediction · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
a lot of the things that are said about consciousness remind me of the questions people were asking themselves in the 18th century or 17th…
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
is how do we get machines to learn predictive models of the world that deal with uncertainty
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I don't even know how to measure an answer to that question.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So whether this will take us all the way to human-level intelligence or something, or just cat-level intelligence is not clear, but among…
Yann LeCun · Not checkable · 2021 |
|
| Unresolvable |
Yann LeCun | Factual |
2021 |
The knowledge needed for house-cat intelligence fits in about 800 million neurons.
Yann LeCun · Factual · 2021 |
Other |
| Not checkable |
Sam Altman | Not checkable |
2021 |
we would no longer motivate people through the fear of not being able to eat.
Sam Altman · Not checkable · 2021 |
|
| Mostly true |
Yann LeCun | Factual |
2021 |
Self-supervised training of visual representations from video has not yet been successful.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not researched |
Yann LeCun | Factual |
2021 |
Representations from self-supervised methods that eliminate augmentation distortions cannot be used for object detection and localization because that information is lost.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So that suggests you only have one model in your head.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
Yann LeCun proposes self-supervised learning as the way to get machines to learn to represent the world.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think those things are actually very simple to learn.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think data augmentation is a temporary necessary evil.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I think we give way too much importance to language as a substrate of intelligence as humans
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
I believe that owning something like a share in America would align all of us in making the country as successful as possible
Sam Altman · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
It sounds very dramatic and very impressive, but it probably isn't mechanistically speaking.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
A sea change in favor of convolutional nets occurred when datasets became bigger and GPUs became available.
Yann LeCun · Factual · 2021 |
AI compute, hardware and data |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
The second reason it may be necessary or at least more efficient is that active learning basically goes for the juggler of what you don't…
Yann LeCun · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Prediction |
2021 |
Automation will significantly change how people think about work.
Sam Altman · Prediction · 2021 |
AI, the economy and work |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
So those things have to be learned.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I'm actually not very positive about
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
Of course this won’t solve all our problems
Sam Altman · Not checkable · 2021 |
|
| Not checkable |
Sam Altman | Not checkable |
2021 |
you really do need money to make money.
Sam Altman · Not checkable · 2021 |
|
| Unresolvable |
Sam Altman | Prediction |
2021 |
If everyone benefits more directly from economic growth, it will be easier to move faster toward a better world with automation.
Sam Altman · Prediction · 2021 |
AI, the economy and work |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
But we judge each other based on our ability
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
I don't get into those discussions because they are kind of pointless.
Yann LeCun · Not checkable · 2021 |
|
| Resolved true |
Yann LeCun | Factual |
2021 |
The American financial payment system is much more lax about signature verification than Europe's.
Yann LeCun · Factual · 2021 |
Economics and markets |
| Unresolvable |
Sam Altman | Prediction |
2021 |
In a world where economic growth guarantees basic needs, absolute poverty in America would be eliminated.
Sam Altman · Prediction · 2021 |
Philosophy, ethics and society |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
And the third one is something we have no idea how to solve.
Yann LeCun · Not checkable · 2021 |
|
| Not checkable |
Yann LeCun | Not checkable |
2021 |
Nature has not made us good at chess.
Yann LeCun · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
Until very recently there were no working methods for training image representations by masking parts of images and reconstructing them.
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not researched |
Yann LeCun | Factual |
2021 |
Oriented edge detectors can be learned within minutes of opening one's eyes.
Yann LeCun · Factual · 2021 |
Science and research methods |
| Not checkable |
Sam Altman | Not checkable |
2021 |
A proposal like this obviously requires a lot of new funding [1] to do at large scale
Sam Altman · Not checkable · 2021 |
|
| Not researched |
Yann LeCun | Factual |
2021 |
LeCun said he believes no text in the world explains the physical situation he was describing (an object that is "sitting on" something else).
Yann LeCun · Factual · 2021 |
AI capabilities |
| Not checkable |
Yann LeCun | Not checkable |
2021 |
But fundamentally, it's still the same thing in building up the intuition about how to, in a self-supprised way to construct background…
Yann LeCun · Not checkable · 2021 |
|