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

DeepMind: A Generalist Agent

deepmind.com

Post reply on HN
Print view View original

191–200 of 345 posts

Re: DeepMind: A Generalist Agent

#191

Post by viksit »

sinenomine wrote: post #185
viksit wrote: post #182

(Former AI researcher / founder here) It always surprises me at the ease at which people jump on a) imminent AGI and b) human extinction in the face of AGI. Would love for someone to correct me / add information here to the contrary. Generalist here just refers to a "multi-faceted agent" vs "General" like AGI. For a) - I see 2 main blockers, 1) A way to build second/third order reasoning systems that rely on intuitio…

It remains to be asked, just why this causal, counterfactual, logical reasoning cannot emerge in a sufficiently scaled-up model trained on a sufficiently diverse real world data? As far as we see, the https://www.gwern.net/Scaling-hypothesis continues to hold, and critics have to move their goalposts every year or two.

Good point. This gets us into the territory of not just "explainable" models, but also the ability to feed into those models "states" in a deterministic way. This is a merger of statistical and symbolic methods in my mind -- and no way for us to achieve this today.

Re: DeepMind: A Generalist Agent

#192

Post by karmasimida »

Would this agent able to handle simple elementary mathematics?

If they are using inspiration from Transformer, then it probably won't be able to count.

For that, I don't really feel that enthusiastic about the 'Generalist' claim, maybe they think this is more catchy than just 'Multi-tasking'?

Re: DeepMind: A Generalist Agent

#193

Post by sinenomine »

chrisco255 wrote: post #35
axg11 wrote: post #22

Slowly but surely we're moving towards general AI. There is a marked split across general society and even ML/AI specialists between those who think that we can achieve AGI using current methods and those who dismiss the possibility. This has always been the case, but what is remarkable about today's environment is that researchers keep making progress contrary to the doubter's predictions. Each time this happens, th…

How do we prepare for super human intelligence? Do you think that the AI will also develop its own motives ? Or will it just be a tool that we're able to plug into and use for ourselves?

We prepare for it by domesticating its lesser forms in practice and searching for ways to increase our own intelligence.

Still, it's pretty likely to end being just a very good intelligent tool, not unlike http://karpathy.github.io/2021/03/27/forward-pass/

Re: DeepMind: A Generalist Agent

#194

Post by kanzure »

fullstackchris wrote: post #173
habitue wrote: post #80

Earlier quoted context omitted.

Agreed, I think if they were to drop the real-time constraint for the sake of the robotics tasks, they could train a huge model with the lessons from PaLM and Chincilla and probably slam dunk the weakly general AI benchmark.

I'm in the camp that thinks we're headed in a perpendicular direction and won't ever get to human levels of AGI with current efforts based on the simple idea that the basic tooling is wrong from first principles. I mean, most of the "progress" in AI has been due to getting better and learning how to understand a single piece of technology: neural networks. A lot of recent neuroscience findings have shown that human b…

> Until we start thinking from the ground up how to build and engineer systems that reflect the human brain, we're essentially wandering around in the dark with perhaps only a piece of what we _think_ is needed for intelligence.

I have wanted an approach based on a top-down architectural view of the human brain. By simulating the different submodules of the human brain (many of which are shared across all animal species), maybe we can make more progress.

https://diyhpl.us/~bryan/papers2/neuro/cognitiveconsilience/...

Machine learning might be a part of the equation at lower levels, although looking at the hippocampus prostheses those only required a few equations:

https://en.wikipedia.org/wiki/Hippocampal_prosthesis#Technol....

Re: DeepMind: A Generalist Agent

#195

Post by mxkopy »

sinenomine wrote: post #185
viksit wrote: post #182

(Former AI researcher / founder here) It always surprises me at the ease at which people jump on a) imminent AGI and b) human extinction in the face of AGI. Would love for someone to correct me / add information here to the contrary. Generalist here just refers to a "multi-faceted agent" vs "General" like AGI. For a) - I see 2 main blockers, 1) A way to build second/third order reasoning systems that rely on intuitio…

It remains to be asked, just why this causal, counterfactual, logical reasoning cannot emerge in a sufficiently scaled-up model trained on a sufficiently diverse real world data? As far as we see, the https://www.gwern.net/Scaling-hypothesis continues to hold, and critics have to move their goalposts every year or two.

Neural networks, at the end of the day, are still advanced forms of data compression. Since they are Turing-complete it is true that given enough data they can learn anything, but only if there is data for it. We haven't solved the problem of reasoning without data, i.e. without learning. The neural network can't, given some new problem that has never appeared in the dataset, in a deterministic way, solve that problem (even given pretrained weights and whatnot). I do think we're pretty close but we haven't come up with the right way of framing the question and combining the tools we have. But I do think the tools are there (optimizing over the space of programs is possible, learning a symbol-space is possible, however symbolic representation is not rigorous or applicable right now)

Re: DeepMind: A Generalist Agent

#196

Post by sinenomine »

viksit wrote: post #191
sinenomine wrote: post #185

Earlier quoted context omitted.

It remains to be asked, just why this causal, counterfactual, logical reasoning cannot emerge in a sufficiently scaled-up model trained on a sufficiently diverse real world data? As far as we see, the https://www.gwern.net/Scaling-hypothesis continues to hold, and critics have to move their goalposts every year or two.

Good point. This gets us into the territory of not just "explainable" models, but also the ability to feed into those models "states" in a deterministic way. This is a merger of statistical and symbolic methods in my mind -- and no way for us to achieve this today.

Why shouldn't we be able to just prompt for it, if our system models natural language well enough?

...

And anyway, this problem of structured knowledge IO has been more or less solved recently: https://arxiv.org/abs/2110.07178

Re: DeepMind: A Generalist Agent

#197

Post by samdjstephens »

idiotsecant wrote: post #188
gurkendoktor wrote: post #179

Earlier quoted context omitted.

Your comment summarizes what I worry might be a more widespread opinion than I expected. If you think that human extinction is a fair price to pay for creating a supercomputer, then our value systems are so incompatible that I really don't know what to say. I guess I wouldn't have been so angry about any of this before I had children, but now I'm very much in favor of prolonged human existence.

> I'm very much in favor of prolonged human existence. Serious question - why?

What are your axioms on what’s important, if not the continued existence of the human race?

edit: I’m genuinely intrigued

Re: DeepMind: A Generalist Agent

#198

Post by croddin »

viksit wrote: post #182

(Former AI researcher / founder here) It always surprises me at the ease at which people jump on a) imminent AGI and b) human extinction in the face of AGI. Would love for someone to correct me / add information here to the contrary. Generalist here just refers to a "multi-faceted agent" vs "General" like AGI. For a) - I see 2 main blockers, 1) A way to build second/third order reasoning systems that rely on intuitio…

I think of it as System 1 vs System 2 thinking from 'Thinking, Fast and Slow' by Daniel Kahneman.[1]

Deep learning is very good at things we can do without thinking, and is in some cases superhuman in those tasks because it can train on so much more data. If you look at the list of tasks in System 1 vs System 2, SOTA Deep learning can do almost everything in System 1 at human or superhuman levels, but not as many in System 2 (although some tasks in System 2 are somewhat ill-defined), System 2 builds on system 1. Sometimes superhuman abilities in System 1 will seem like System 2. (A chess master can beat a noob without thinking while the noob might be thinking really hard. Also GPT-3 probably knows 2+2=4 from training data but not 17 * 24, although maybe with more training data it would be able to do math with more digits 'without thinking' ).

System 1 is basically solved, but System 2 is not. System 2 could be close behind System 2 by building on System 1 but it isn't clear how long that will take.

[1]. https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow#Summar...

Re: DeepMind: A Generalist Agent

#199

Post by TaupeRanger »

robitsT6 wrote: post #107
TaupeRanger wrote: post #97

Earlier quoted context omitted.

And yet, the only thing that really matters out of your entire list is the 1st one: that AI solves problems that actually improve the human condition. And Alpha Fold has not done that at all. It may be very nice for people interested in protein folding, but until it actually helps us find something that we wouldn't have found otherwise, and that discovery leads to (for example) an ACTUAL drug or treatment that helps…

But there have been quite a few scientific papers that have used discoveries from AlphaFold already. There have been many scientists who have been stuck for years, who are suddenly past their previous bottlenecks. What gives you the impression that it hasn't helped us?

I am not saying that Alpha Fold won't help scientists publish papers. I am just skeptical (though still hopeful) of it doing anything to improve the human condition by actually making human existence better. Publishing papers can be of neutral or negative utility in that realm.

Re: DeepMind: A Generalist Agent

#200

Post by sinenomine »

mxkopy wrote: post #195
sinenomine wrote: post #185

Earlier quoted context omitted.

It remains to be asked, just why this causal, counterfactual, logical reasoning cannot emerge in a sufficiently scaled-up model trained on a sufficiently diverse real world data? As far as we see, the https://www.gwern.net/Scaling-hypothesis continues to hold, and critics have to move their goalposts every year or two.

Neural networks, at the end of the day, are still advanced forms of data compression. Since they are Turing-complete it is true that given enough data they can learn anything, but only if there is data for it. We haven't solved the problem of reasoning without data, i.e. without learning. The neural network can't, given some new problem that has never appeared in the dataset, in a deterministic way, solve that proble…

I do think we underestimate compressionism[1] especially in the practically achievable limit.

Sequence prediction is closely related to optimal compression, and both basically require the system to model the ever wider context of the "data generation process" in ever finer detail. In the limit this process has to start computing some close enough approximation of the largest data-generating domains known to us - history, societies and persons, discourse and ideas, perhaps even some shadow of our physical reality.

In the practical limit it should boil down to exquisite modeling of the person prompting the AI to do X given the minimum amount of data possible. Perhaps even that X you had in mind when you wrote your comment.

1. http://ceur-ws.org/Vol-1419/paper0045.pdf

Post reply on HN