Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

venturebeat.com

Post reply on HN
Print view View original

51–60 of 181 posts

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#51

Post by kahoon »

karmasimida wrote: post #17

And no one should be surprised by this. The NN advancement of late doesn't help addressing human-style symbolic reasoning at all. All we have is a much more powerful function approximator with a drastic increased capacity (very deep networks with billions of parameters) and scalable training scheme (SGD and its variants). Such architecture works great for differentiable data, such's images/audios, but the improvement…

I don't understand this fixation on symbolic reasoning. Do any other animals practice this? If the answer is no, then it is probably not the most important milestone to AGI or at least not the one we should be currently aiming for. Right now we can not replicate the cognition of a mouse. Feels like we want to go to Mars before figuring out how to build a rocket.

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#52

Post by darkmighty »

toasterlovin wrote: post #5

Not to mention that we don't even know if general intelligence exists. All we know is that mental abilities tend to correlate, but not why they tend to correlate. And if you think about designing machines, in general, the idea of general intelligence is utterly ridiculous. Does a fast car have general speediness? Of course not, it has dozens or hundreds of discrete optimizations that all contribute in some degree to…

Well, we have general purpose processors. You can prove they can run any algorithm you want (i.e. are Turing complete), but also, for practical problems (i.e. the ones encountered in engineering solutions in our planet and in our universe), they give reasonable max-min performance. Analogously I don't think 'AGI' is entirely useless -- you'd expect an AGI to have some properties like being able to solve reasonably well problems found in nature and society, maybe have a motivational framework distinguishing it as a separate entity, some knowledge about the world, etc.

edit: In terms of Turing-completeness analogues, the best candidate for AGI I think would be simply brute force capability: can this agent try all possible solutions until it solves this problem? (obviously using a heuristic to prioritize) -- that is, it'd employ a form of Universal Search[1] (aka Levin Search). Humans don't necessarily pass this test rigorously because we'd always get bored with a problem and because we have finite memory. But then CPUs are not truly Turing complete either (it's "just" a good model).

[1] http://www.scholarpedia.org/article/Universal_search

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#53

Post by red75prime »

yters wrote: post #12
siekmanj wrote: post #7

Earlier quoted context omitted.

We know the brain is doing something - if you don't want to call it computation, then you might as well call it magic.

There are other possibilities. For example, there can be an immaterial mind that operates as a halting oracle and interfaces with the world through the brain. Halting oracles are well defined, and we can empirically test for their existence. So, no reason why we have to assume everything humans do is reducible to some sort of automata. The only reason we make the assumption is because of prior materialistic commitmen…

> We programmers write halting programs with great regularity.

Making any program halting program is trivial: add executed instructions counter, halt program at some value of the counter. Proving that an arbitrary program halts is an entirely different task.

> So, the fact that we cannot solve some problems does not imply we are not halting oracles.

If it's allowed to not solve some problems, then I can write such an oracle:

Run a program for a million steps. If program has halted, output "Halts", otherwise output "Don't know".

It can't solve some problems, but by your logic it doesn't imply it's not a halting oracle. You are missing something.

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#54

Post by mindgam3 »

Demis Hassabis (true) statements here would be much more credible if DeepMind wasn't currently making a mint by promoting AlphaZero to the masses as a "general purpose artificial intelligence system".

Don't believe me? Check out this series of marketing videos on YouTube by GM Matthew Sadler.

1. “Hi, I’m GM Matthew Sadler, and in this series of videos we’re taking a look at new games between AlphaZero, DeepMind’s general purpose artificial intelligence system, and Stockfish” (1)

2. “Hi, I’m GM Matthew Sadler, and welcome to this review of the World Champinship match between Magnus Carlsen and Fabiano Caruana. And it’s a review with a difference, because we are taking a look at the games together with AlphaZero, DeepMind’s general purpose artificial intelligence system...” (2)

3. “Hi, I’m GM Matthew Sadler, and in this video we’ll be taking a look at a game between AlphaZero, DeepMind’s general purpose artificial intelligence system, and Stockfish” (3)

I could go on, but you get my point. Search youtube for "Sadler DeepMind" and you'll see all the rest. This is a script.

But wait, you say, that's just some random unaffiliated independent grandmaster who just happens to be using an inaccurate script on his own, no DeepMind connection at all! And to that I would say, check out this same random GM being quoted directly on DeepMind's blog waxing eloquently and rapturously about AlphaZero's incredible qualities. (4)

Let's be clear. I am in no way dismissing AlphaZero's truy remarkable abilities in both chess and other games like go and shogi. Nor do I have a problem with Demis Hassabis making headlines for stating the obvious about deep learning (that it's good at solving certain limited types of puzzles, but we are a long way from AGI, why is this controversial).

My problem is that Hassabis is speaking out of both sides of his mouth. Increasing DeepMind/Google's value by many millions with his marketing message, while acting like he's not doing that. It feels intellectually dishonest.

To solve this, all DeepMind needs to stop instructing its Grandmaster mouthpieces to refer to AlphaZero as a "general articial intelligence system". Let's see how long that takes.

(1) https://www.youtube.com/watch?v=2-wFUdvKTVQ&t=0m10s (2) https://www.youtube.com/watch?v=X4T0_IoGQCE&t=0m05s (3) https://www.youtube.com/watch?v=jS26Ct34YrQ&t=0m05s (4) https://deepmind.com/blog/alphazero-shedding-new-light-grand...

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#55

Post by joe_the_user »

skybrian wrote: post #29
karmasimida wrote: post #17

And no one should be surprised by this. The NN advancement of late doesn't help addressing human-style symbolic reasoning at all. All we have is a much more powerful function approximator with a drastic increased capacity (very deep networks with billions of parameters) and scalable training scheme (SGD and its variants). Such architecture works great for differentiable data, such's images/audios, but the improvement…

Hmm, it seems like natural language translation has been getting quite a bit better with statistical techniques, though? I guess it depends what you mean by "only incremental".

Whenever I attempt to use a translate site to translate more than a paragraph (Facebook or Google), it comes out a garbled mess - doesn't mean some sentences are seemingly clear and meaningful. The big thing is it chokes on idioms, not understanding, not leaving them as is but guessing some clearly wrong meaning. I occasionally find single-sentence posts on Facebook apparently translated astonishingly well, in the sense of being literate English and seeming to reflect the original meaning. But my French or Spanish is quite rough so translate could have missed something big - as I know it does when you get into longer texts.

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#56

Post by joe_the_user »

hacker_9 wrote: post #8

Behind every successful neural network is a human brain. Neural networks are a tool, an advanced tool for sure, but still just a tool. If we are looking for AGI, and assuming the brain is an AGI, then there are still many differences to resolve. For example, back propagation has not been observed in nature. Nor has gradient descent. So the core mechanisms for learning in nature have still to reveal their secrets.

Behind every successful brain is a little strand of DNA and some environmental inputs. Somehow a brain might be more than the DNA however.

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#57

Post by qqqwww »

yters wrote: post #12
siekmanj wrote: post #7

Earlier quoted context omitted.

We know the brain is doing something - if you don't want to call it computation, then you might as well call it magic.

There are other possibilities. For example, there can be an immaterial mind that operates as a halting oracle and interfaces with the world through the brain. Halting oracles are well defined, and we can empirically test for their existence. So, no reason why we have to assume everything humans do is reducible to some sort of automata. The only reason we make the assumption is because of prior materialistic commitmen…

> A Turing machine has to be given the axiom of infinity to make this kind of inference, it cannot derive it in any way.

Why not? Are you aware of a proof of this? I think you are limiting the capabilities of Turing machines without evidence.

> Unlike the game AIs that repeatedly try to walk through walls.

Game AIs capabilities are a small subset of what a Turing machine can do. Most game AIs can't do speech recognition or solve math equations either.

> We programmers write halting programs with great regularity.

So do other programs. Writing a halting program is not an uncomputable problem, and doesn't require solving the halting problem.

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#58

Post by darawk »

karmasimida wrote: post #17

And no one should be surprised by this. The NN advancement of late doesn't help addressing human-style symbolic reasoning at all. All we have is a much more powerful function approximator with a drastic increased capacity (very deep networks with billions of parameters) and scalable training scheme (SGD and its variants). Such architecture works great for differentiable data, such's images/audios, but the improvement…

> And no one should be surprised by this. The NN advancement of late doesn't help addressing human-style symbolic reasoning at all. All we have is a much more powerful function approximator with a drastic increased capacity (very deep networks with billions of parameters) and scalable training scheme (SGD and its variants).

You think symbolic reasoning is not a function? In what sense do you think 'symbolic reasoning' is a distinct thing from 'function approximation'?

Re: Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality

#59

Post by atschantz »

hacker_9 wrote: post #8

Behind every successful neural network is a human brain. Neural networks are a tool, an advanced tool for sure, but still just a tool. If we are looking for AGI, and assuming the brain is an AGI, then there are still many differences to resolve. For example, back propagation has not been observed in nature. Nor has gradient descent. So the core mechanisms for learning in nature have still to reveal their secrets.

I mean it's difficult to 'observe' gradient descent, there are no characteristic properties that you can identify without specifying the relative objective function. But most of the process theories from computational neuroscience are based on some form of gradient descent. Even if it's only implicit, you'll be able to describe the variables of the system as moving against the gradient of some function.

But yes, it's extremely unlikely that nature implements backpropagation directly, as it relies on non-local gradients.

Post reply on HN