DeepMind says reinforcement learning is ‘enough’ to reach general AI • Hacker News Board

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#91

Post by TeMPOraL » Thu, Jun 10, 2021, 8:17 AM UTC

otabdeveloper4 wrote: post #47

harry8 wrote: post #18

Some Bozo who has heard all this many times before is suspicious of claims from places like Deep Mind who have a financial incentive to make them (keep funding) where there aren't working machines to back that claim up. Some Bozo has no credentials, no reputation, no track record of publications and barely supports the claim they're making with anything much. Some Bozo has no financial incentives or otherwise to opin…

I'll bet a sum of real money that Some Bozo is correct. Bozo has the hindsight of history and philosophy going for him, while Deep Mind has a huge financial temptation to sell snake oil.

I'll happily bet fake internet points instead:

https://www.metaculus.com/questions/create/

EDIT: bunch of other related predictions currently open:

https://www.metaculus.com/questions/?order_by=-activity&sear...

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#92

Post by anyfoo » Thu, Jun 10, 2021, 8:21 AM UTC

gammalost wrote: post #88

anyfoo wrote: post #73

Earlier quoted context omitted.

> We know that Turing machines are very limited things and that the computational processes they carry out also very limited in applicability. Again, source? Do you know of anything that is able to perform a computation that a Turing machine cannot? > Just the fact that we can imagine things that can't be computed by a Turing machine should clue you in that it's probably otherwise. Like what? Uncomputable numbers lik…

>Like what? The halting problem is (I think?) the standard example. Here's a short video on it https://youtu.be/macM_MtS_w4

Chaitin's constant is closely tied to the halting problem: Each bit essentially tells you if a program (in the particular order the programs are listed in the constant) halts or not. Computing every bit in there would mean knowing for every program if it halts or not, which would solve the halting problem, which we can prove is impossible. So we cannot compute that constant.

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#93

Post by croon » Thu, Jun 10, 2021, 8:22 AM UTC

tinco wrote: post #80

0xB31B1B wrote: post #51

Earlier quoted context omitted.

Protein folding is a well modeled math problem. The alpha fold solution is extremely good at pruning (aka guessing) folding chain structure possibilities. I am Impressed and this is a difficult problem but this is extremely different from AGI as this is a well scoped easily modelable problem that is basically a chain of 26 inputs types of links of arbitrary length. I am not trying to take away that the protein foldin…

It is entirely not necessary for an AGI to be able to drive a car. Frankly, after seeing AlphaZero and AlphaFold I'm surprised they didn't declare AGI right there and then. People assume that when AGI happens, computers can suddenly outsmart humans in every way and solve every problem imaginable. The reality is just that it could in theory given enough time and resources. It is like quantum computing. In theory it ca…

> It is entirely not necessary for an AGI to be able to drive a car.

"Artificial general intelligence (AGI) is the hypothetical[1] ability of an intelligent agent to understand or learn any intellectual task that a human being can."

What definition are you using?

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#94

Post by zimpenfish » Thu, Jun 10, 2021, 8:22 AM UTC

d110af5ccf wrote: post #78

rich_sasha wrote: post #64

Earlier quoted context omitted.

Is Deep Mind a “booming business”? They are achieving great things academically, but their business successes are either kept secret or mostly absent. All I know about is the Google data centre cooling scheduling, probably a big saving for Google but hardly an achievement that in its own professes their business success.

Deep Mind is cutting edge ML in general, right? Doesn't Google actively apply the lessons learned all over the place? YouTube content recommendation stands out to me in particular. Translation and automated closed captioning are also obviously ML based. I'd guess that most of the really interesting stuff would be behind the scenes and not immediately visible to end users though.

> YouTube content recommendation stands out to me in particular.

If that's "cutting edge ML", then going off my YouTube recommendations, we're back in another AI winter. If I watch one video from a channel I've not seen before, I'll get that channel recommended constantly even if it bears no resemblance to what I normally watch. On my Explore page, the first 22 videos (of which 8 are Fortnite-related!) hold no interest for me. My Home page is just channels I've watched repeatedly and/or am subscribed to. It's a mess.

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#95

Post by graderjs » Thu, Jun 10, 2021, 8:23 AM UTC

harry8 wrote: post #18

Some Bozo who has heard all this many times before is suspicious of claims from places like Deep Mind who have a financial incentive to make them (keep funding) where there aren't working machines to back that claim up. Some Bozo has no credentials, no reputation, no track record of publications and barely supports the claim they're making with anything much. Some Bozo has no financial incentives or otherwise to opin…

Plot-twist: Deep Mind comes out as Some Bozo comment author

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#96

Post by ausbah » Thu, Jun 10, 2021, 8:32 AM UTC

graderjs wrote: post #95

harry8 wrote: post #18

Some Bozo who has heard all this many times before is suspicious of claims from places like Deep Mind who have a financial incentive to make them (keep funding) where there aren't working machines to back that claim up. Some Bozo has no credentials, no reputation, no track record of publications and barely supports the claim they're making with anything much. Some Bozo has no financial incentives or otherwise to opin…

Plot-twist: Deep Mind comes out as Some Bozo comment author

Bozo et al

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#97

Post by ausbah » Thu, Jun 10, 2021, 8:33 AM UTC

A-Train wrote: post #76

rusk wrote: post #42

Earlier quoted context omitted.

Edit: sorry just realised you’re making the same point as me more or less. Putting yourself in third person. I’ll let my comment stand anyhow :) Screwing my face up, looking at this sideways … but it seems as though you’re saying that the Bozos of HN have nothing useful to contribute to this discussion based on … [rereads] … their lack of academic credentials in the area… you could say this about just about any HN po…

The algorithms to train, initialize the networks, new architectures are far more important than the hardware advances. If people knew how to train NNs 50 years ago we would live in a different world.

the basic concepts underlying DNNs have been known for decades, it has been exponential increases in compute power that have made them practical

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#98

Post by otabdeveloper4 » Thu, Jun 10, 2021, 8:34 AM UTC

anyfoo wrote: post #73

otabdeveloper4 wrote: post #71

Earlier quoted context omitted.

Assuming that the universe can be simulated by a Turing machine is a strong and weird claim that needs to be defended, not the other way around. We know that Turing machines are very limited things and that the computational processes they carry out also very limited in applicability. What's the evidence that the universe is more limited that a Turing machine? Just the fact that we can imagine things that can't be co…

> We know that Turing machines are very limited things and that the computational processes they carry out also very limited in applicability. Again, source? Do you know of anything that is able to perform a computation that a Turing machine cannot? > Just the fact that we can imagine things that can't be computed by a Turing machine should clue you in that it's probably otherwise. Like what? Uncomputable numbers lik…

> Do you know of anything that is able to perform a computation that a Turing machine cannot?

That's a circular argument, because "computation" is literally defined as "something that can be computed by a Turing machine".

That said, the first month of the first year of a CS education is "here's these problems that can't be solved by a Turing machine, mind=blown". (At least where I studied CS, that is.)

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#99

Post by marcescence » Thu, Jun 10, 2021, 8:36 AM UTC

TeMPOraL wrote: post #91

otabdeveloper4 wrote: post #47

Earlier quoted context omitted.

I'll bet a sum of real money that Some Bozo is correct. Bozo has the hindsight of history and philosophy going for him, while Deep Mind has a huge financial temptation to sell snake oil.

I'll happily bet fake internet points instead: https://www.metaculus.com/questions/create/ EDIT: bunch of other related predictions currently open: https://www.metaculus.com/questions/?order_by=-activity&sear... https://www.metaculus.com/questions/?order_by=-activity&sear...

A quote from datscilly, the top forecaster on metaculus:

>AGI may never happen, but the chance of that is small enough that adjusting for that here will not make a big difference (I put ~10% that AGI will not happen for 500 years or more, but it already matches that distribution quite well).[1]

[1]:https://www.lesswrong.com/posts/hQysqfSEzciRazx8k/forecastin...

Re: DeepMind says reinforcement learning is ‘enough’ to reach general AI

#100

Post by qsort » Thu, Jun 10, 2021, 8:38 AM UTC

tinco wrote: post #80

0xB31B1B wrote: post #51

Earlier quoted context omitted.

Protein folding is a well modeled math problem. The alpha fold solution is extremely good at pruning (aka guessing) folding chain structure possibilities. I am Impressed and this is a difficult problem but this is extremely different from AGI as this is a well scoped easily modelable problem that is basically a chain of 26 inputs types of links of arbitrary length. I am not trying to take away that the protein foldin…

It is entirely not necessary for an AGI to be able to drive a car. Frankly, after seeing AlphaZero and AlphaFold I'm surprised they didn't declare AGI right there and then. People assume that when AGI happens, computers can suddenly outsmart humans in every way and solve every problem imaginable. The reality is just that it could in theory given enough time and resources. It is like quantum computing. In theory it ca…

> it could in theory given enough time and resources.

In theory given enough time and resources, anyone can defeat any grandmaster in Chess: just compute the extended tree form of the game and run the minimax algorithm.

The "given enough time and resources" clause makes everything that follows meaningless, unless a reasonable algorithm is presented.

> It is like quantum computing.

It is absolutely not like quantum computing. Shor's algorithm is something you can look up right now. It is precise and well-defined. The problems we are facing with quantum computation are related to the fact that we can't really build reliable hardware. But we know that given such machines the algorithm would work. We have precise bounds and requirements on those machines.

As far as AGI goes, we have absolutely no idea. There's lively debate on whether anything we have done even counts as significant advancement towards AGI.