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

Artificial General Intelligence – A gentle introduction

cis.temple.edu

Post reply on HN
Print view View original

191–200 of 201 posts

Re: Artificial General Intelligence – A gentle introduction

#191

Post by famouswaffles »

charcircuit wrote: post #183
famouswaffles wrote: post #118

Earlier quoted context omitted.

Nobody said anything about Magic. >There is no magic here at the lowest level – you can interrogate the math at each step and it'll make sense. See that's the thing. You can't unless "making sense" has lost all meaning. That you can see a bunch of signals firing or matrices being multiplied does not mean they "make sense" or are meaningful to you. Lol level gibberish is still gibberish. Our ability to divine the purp…

>Our ability to divine the purpose of activations of anything but the extremely small scale is atrocious. The value of each parameter is chosen to minimize the loss. This applies to every single weight of the model. Not all weighs affect loss the same amount which is why concepts like pruning exist.

>The value of each parameter is chosen to minimize the loss

Vague and fairly useless. What is it doing to minimize loss ?

>Not all weighs affect loss the same amount which is why concepts like pruning exist.

Only weights with values close to or at zero get pruned. It's not because we know what each weight does and can tell what would work otherwise.

Re: Artificial General Intelligence – A gentle introduction

#192

Post by charcircuit »

famouswaffles wrote: post #191
charcircuit wrote: post #183

Earlier quoted context omitted.

>Our ability to divine the purpose of activations of anything but the extremely small scale is atrocious. The value of each parameter is chosen to minimize the loss. This applies to every single weight of the model. Not all weighs affect loss the same amount which is why concepts like pruning exist.

>The value of each parameter is chosen to minimize the loss Vague and fairly useless. What is it doing to minimize loss ? >Not all weighs affect loss the same amount which is why concepts like pruning exist. Only weights with values close to or at zero get pruned. It's not because we know what each weight does and can tell what would work otherwise.

>Vague and fairly useless.

When creating a model your goal is to find one with minimal loss. Being able to figure how to improve a model by finding weights that reduce the loss is not a vague or useless idea.

>What is it doing to minimize loss?

The value helps us get to a location in the parameter space with lower loss.

>Only weights with values close to or at zero get pruned.

Weights near 0 don't change the results of the calculations they are used in my much which is why they don't effect loss very much.

Re: Artificial General Intelligence – A gentle introduction

#193

Post by cubefox »

theptip wrote: post #143
cubefox wrote: post #122

Earlier quoted context omitted.

AGI us not a very useful term, because people use it often synonymous with "human level or higher ability". But the opposite of "general AI" is not "less intelligent than a human" but "narrow AI". The narrow/general distinction is orthogonal to the low/high ability ("intelligence") distinction. All animals are very general, as their domain of operation is the real world, not some narrow modality like strings of text…

Personally I take AGI to refer to a system that is both “intelligent enough” and “general enough”. Given the existence of super-human narrow AI, the interesting property is generality, not intelligence. But I don’t think it’s useful to call a sub-human cat-level general AI an AGI. Some would disagree; there was a paper arguing that ChatGPT is weak AGI. But as I see it AGI is a term of art that refers to a point on th…

Given the existence of super-human narrow AI, the interesting property is generality, not intelligence. But I don’t think it’s useful to call a sub-human cat-level general AI an AGI.

If we have AI as general as an animal, ASI (superintelligence) is probably imminent. Because the architecture of humans intelligence probably isn't very different from cats, just the scale is bigger.

Re: Artificial General Intelligence – A gentle introduction

#194

Post by famouswaffles »

charcircuit wrote: post #192
famouswaffles wrote: post #191

Earlier quoted context omitted.

>The value of each parameter is chosen to minimize the loss Vague and fairly useless. What is it doing to minimize loss ? >Not all weighs affect loss the same amount which is why concepts like pruning exist. Only weights with values close to or at zero get pruned. It's not because we know what each weight does and can tell what would work otherwise.

>Vague and fairly useless. When creating a model your goal is to find one with minimal loss. Being able to figure how to improve a model by finding weights that reduce the loss is not a vague or useless idea. >What is it doing to minimize loss? The value helps us get to a location in the parameter space with lower loss. >Only weights with values close to or at zero get pruned. Weights near 0 don't change the results…

>When creating a model your goal is to find one with minimal loss. Being able to figure how to improve a model by finding weights that reduce the loss is not a vague or useless idea.

I'm sorry but did you bother reading the previous conversation ? We were talking about how much we know what weights do during inference. "It reduces loss" alone is in fact very vague and useless for interpretability.

>The value helps us get to a location in the parameter space with lower loss.

What neuron(s) is responsible for capitalization in GPT? You wouldn't get that simply from "reduces the loss". Our understanding of what the neurons do is very limited.

>Weights near 0 don't change the results of the calculations they are used in my much which is why they don't effect loss very much.

I understand that lol.

"This value is literally 0 so it can't affect things much" is a very different understanding level from "this bunch of weights are a redundancy because this set already achieves this function that this other set does and so can be pruned. Let's also tune this set so it never tries to call this other set while we're at it. "

Re: Artificial General Intelligence – A gentle introduction

#195

Post by gondolgames »

trabant00 wrote: post #56
tpoacher wrote: post #49

AGI is not a difficult "problem". It's a difficult "definition". Given very specific, practical, functional definitions, AGI is a breeze.

I keep hearing this argument that we already have AI, it's just the naysayers move the definition. Level 5 autonomous driving is specific, practical, functionally defined. Where's the breeze implementation?

Are you sure level 5 autonomous driving is specific? What would be the exact goal behind decisions in such a system? Not even talking about the trolley problem, would the software optimize for speed or not harming people, for example? Obviously we would want a combination of both, otherwise people can either get harmed or not get anywhere in time. But then, how fast should it take a corner? How much chance of human harm should it allow to get somewhere fast? Furthermore, what do we mean by human harm? The system would obviously need to know what human harm is, to be able to avoid it. Which requires defining human and defining harm, both of which are incredibly difficult to do specifically - more on this by Rob Miles here: https://www.youtube.com/watch?v=7PKx3kS7f4A I don't think level 5 autonomous driving is specifically defined. We just don't have systems intelligent enough for this to be a problem yet.

Re: Artificial General Intelligence – A gentle introduction

#196

Post by sdenton4 »

YeGoblynQueenne wrote: post #189
sdenton4 wrote: post #169

Earlier quoted context omitted.

The 'game playing' you lead with is clearly dominated by deep learning in this decade.

When I wrote the comment you replied to I was thinking specifically, and, admittedly narrowly, of adversarial search rather than general game playing but even so it's not that simple. Deep Learning is certainly dominant in computer games like Atari. However, in classic board games dominant systems combine deep learning and classical search-based approaches (namely Monte-Carlo Tree Search, MCTS, a stochastic version o…

I'm very familiar with how mcts is used in alpha go and mu zero.

I'm not sure how you can say it's hidden in the details: the name of the paper is "mastering go with deep neutral networks and tree search."

It's also not an oversell on the deep learning component. Per the ablations in the alpha go paper, the no-mcts ELO is over 2000, while the mcts-only ELO is a bit under 1500. Combining the two gives an ELO of nearly 3000. So the deep learning system is outperforming the mcts-only system, and gets a significant boost from using mcts.

The mu zero paper also does not hide the tree search; it is prominent in the figures and mentioned in captions, for example. It is not the main focus of the paper, though, so perhaps isn't discussed as much as in the alpha go paper.

(Weirdly axe-grindy comment...)

Re: Artificial General Intelligence – A gentle introduction

#197

Post by charcircuit »

famouswaffles wrote: post #194
charcircuit wrote: post #192

Earlier quoted context omitted.

>Vague and fairly useless. When creating a model your goal is to find one with minimal loss. Being able to figure how to improve a model by finding weights that reduce the loss is not a vague or useless idea. >What is it doing to minimize loss? The value helps us get to a location in the parameter space with lower loss. >Only weights with values close to or at zero get pruned. Weights near 0 don't change the results…

>When creating a model your goal is to find one with minimal loss. Being able to figure how to improve a model by finding weights that reduce the loss is not a vague or useless idea. I'm sorry but did you bother reading the previous conversation ? We were talking about how much we know what weights do during inference. "It reduces loss" alone is in fact very vague and useless for interpretability. >The value helps us…

>What neuron(s) is responsible for capitalization in GPT?

It doesn't matter. Individual things like capitalization are vague and useless for interpretability. We know that incorrect capitalization will increase loss, so the model will need to figure how to do it correctly.

>Our understanding of what the neurons do is very limited.

The mathematical definition is right in the code. You can see the calculations they are doing.

>this bunch of weights are a redundancy because this set already achieves this function that this other set does and so can be pruned. Let's also tune this set so it never tries to call this other set while we're at it.

They are equivalent. If removing something does not increase loss then it was redundant behavior at least for the dataset that it is being tested against.

Re: Artificial General Intelligence – A gentle introduction

#198

Post by theptip »

cubefox wrote: post #193
theptip wrote: post #143

Earlier quoted context omitted.

Personally I take AGI to refer to a system that is both “intelligent enough” and “general enough”. Given the existence of super-human narrow AI, the interesting property is generality, not intelligence. But I don’t think it’s useful to call a sub-human cat-level general AI an AGI. Some would disagree; there was a paper arguing that ChatGPT is weak AGI. But as I see it AGI is a term of art that refers to a point on th…

Given the existence of super-human narrow AI, the interesting property is generality, not intelligence. But I don’t think it’s useful to call a sub-human cat-level general AI an AGI. If we have AI as general as an animal, ASI (superintelligence) is probably imminent. Because the architecture of humans intelligence probably isn't very different from cats, just the scale is bigger.

I think that very well could be true, depends on how that generality was obtained.

I would not be surprised if a multi-modal LLM (basically current architecture) could be wired up to be as general as a cat with current param count, and with the spark of human creativity (AGI/ASI) still ending up being far away.

But if you made a new architecture that solved the generalization problem (ie baking in a world model, self-symbol, etc) but only reached cat intelligence, then it would seem very likely that human-level was soon to follow.

Re: Artificial General Intelligence – A gentle introduction

#199

Post by YeGoblynQueenne »

sdenton4 wrote: post #196
YeGoblynQueenne wrote: post #189

Earlier quoted context omitted.

When I wrote the comment you replied to I was thinking specifically, and, admittedly narrowly, of adversarial search rather than general game playing but even so it's not that simple. Deep Learning is certainly dominant in computer games like Atari. However, in classic board games dominant systems combine deep learning and classical search-based approaches (namely Monte-Carlo Tree Search, MCTS, a stochastic version o…

I'm very familiar with how mcts is used in alpha go and mu zero. I'm not sure how you can say it's hidden in the details: the name of the paper is "mastering go with deep neutral networks and tree search." It's also not an oversell on the deep learning component. Per the ablations in the alpha go paper, the no-mcts ELO is over 2000, while the mcts-only ELO is a bit under 1500. Combining the two gives an ELO of nearly…

Well I haven't read those papers since they came out so I will defer to your evidently better recollection. It seems I formed an impression from what was going around on HN and the media at the time and I misremember the content of the papers.

>> (Weirdly axe-grindy comment...)

https://youtu.be/m9KbmRTgigQ

Re: Artificial General Intelligence – A gentle introduction

#200

Post by wizofaus »

janalsncm wrote: post #174
wizofaus wrote: post #173

Earlier quoted context omitted.

And once have an AI engine that, given just the instructions on how to drive a car and a list of road rules, can operate one perfectly, I'd agree we're a huge step closer to an AGI (if it can also learn how to do all the other things most humans can just given similar inputs, then sure, it would qualify unreservedly).

Sure, and at that point we can shift the goalposts to some other task since driving (like chess) will seem easy in retrospect. Put another way, what would a system which has taught itself to drive tell us about general intelligence that we didn’t already know? Because as of now it seems like the pattern is Computers could never do X Computers can’t do X Computers can’t do X very well Computers can’t do X well in some…

> Sure, and at that point we can shift the goalposts to some other task

If you like. But I'm happy with where I have them. I'm also pretty confident I'll see that goal reached in my lifetime.

Post reply on HN