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

AGI is an engineering problem, not a model training problem

vincirufus.com

Post reply on HN
Print view View original

131–140 of 442 posts

Re: AGI is an engineering problem, not a model training problem

#131

Post by jibal »

glitchc wrote: post #10

We don't know if AGI is even possible outside of a biological construct yet. This is key. Can we land on AGI without some clear indication of possibility (aka Chappie style)? Possibly, but the likelihood is low. Quite low. It's essentially groping in the dark. A good contrast is quantum computing. We know that's possible, even feasible, and now are trying to overcome the engineering hurdles. And people still think th…

It's not "key"; it's not even relevant ... the proof will be in the pudding. Proving a priori that some outcome is possible plays no role in achieving it. And you slid, motte-and-bailey-like, from "know" to "some clear indication of possibility" -- we have extremely clear indications that it's possible, since there's no reason other than a belief in magic to think that "biological" is a necessity.

Whether is feasible or practical or desirable to achieve AGI is another matter, but the OP lays out multiple problem areas to tackle.

Re: AGI is an engineering problem, not a model training problem

#132

Post by supermatt »

energy123 wrote: post #126
andy99 wrote: post #19

If you believe the bitter lesson, all the handwavy "engineering" is better done with more data. Someone likely would have written the same thing as this 8 years ago about what it would take to get current LLM performance. So I don't buy the engineering angle, I also don't think LLMs will scale up to AGI as imagined by Asimov or any of the usual sci-fi tropes. There is something more fundamental missing, as in missing…

What will it scale up to if not AGI? OpenAI has a synthetic data flywheel. What are the asymptotics of this flywheel assuming no qualitative additional breakthrough?

What will shouting louder achieve if not wisdom?

Re: AGI is an engineering problem, not a model training problem

#133

Post by Davidzheng »

walleeee wrote: post #88
AIPedant wrote: post #56

Earlier quoted context omitted.

It is vacuously true that a Turing machine can implement human intelligence: simply solve the Schrödinger equation for every atom in the human body and local environment. Obviously this is cost-prohibitive and we don’t have even 0.1% of the data required to make the simulation. Maybe we could simulate every single neuron instead, but again it’ll take many decades to gather the data in living human brains, and it woul…

> It is vacuously true that a Turing machine can implement human intelligence The case of simulating all known physics is stronger so I'll consider that. But still it tells us nothing, as the Turing machine can't be built. It is a kind of tautology wherein computation is taken to "run" the universe via the formalism of quantum mechanics, which is taken to be a complete description of reality, permitting the assumptio…

QM is a testable hypothesis, so I don't think it's necessarily like an axiomatic assumption here. I'm not sure what you mean by "it tells us nothing, as ... can't be built". It tells us there's no theoretical constraint and only an engineering constraint to doing simulating the human brain (and all the tasks)

Re: AGI is an engineering problem, not a model training problem

#134

Post by bsenftner »

dragonwriter wrote: post #128
bsenftner wrote: post #125

AGI, by definition, in its name Artificial General Intelligence implies / directly states that this type of AI is not some dumb AI that requires training for all its knowledge, a general intelligence merely needs to be taught how to count, the basic rules of logic, and the basic rules of a single human language. From those basics all derivable logical human sciences will be rediscovered by that AGI and our next job i…

> AGI, by definition, in its name Artificial General Intelligence implies / directly states that this type of AI is not some dumb AI that requires training for all its knowledge, a general intelligence merely needs to be taught how to count, the basic rules of logic, and the basic rules of a single human language. From those basics all derivable logical human sciences will be rediscovered by that AGI That's not how n…

Are you sure? Do you require dozens, to hundreds, to thousands of examples before you understand a concept? I expect no. That is because you have comprehension that can generalize a situation to basic concepts which you apply to other situations without effort. You comprehend. AI cannot do that: get the idea from a few, under a half dozen examples if necessary. Often a human needs 1-3 examples before they can generalize any concept. Not AI.

Re: AGI is an engineering problem, not a model training problem

#135

Post by dsubburam »

bee_rider wrote: post #63
mattnewton wrote: post #43

Earlier quoted context omitted.

Why does it need to exclude all non human animals? Could it not be a difference of degree rather than of kind?

The post I was responding to had > On the contrary, we have one working example of general intelligence (humans) I think some animals probably have what most people would informally call general intelligence, but maybe there’s some technical definition that makes me wrong.

I do not know how "general intelligence" is defined, but there are a set of features we humans have that other animals mostly don't, as per the philosopher Roger Scruton[1], that I am reproducing from memory (errors mine):

1. Animals have desires, but do not make choices

We can choose to do what we do not desire, and choose not to do what we desire. For animals, one does not need to make this distinction to explain their behavior (Occam's razor)--they simply do what they desire.

2. Animals "live in a world of perception" (Schopenhauer)

They only engage with things as they are. They do not reminisce about the past, plan for the future, or fantasize about the impossible. They do not ask "what if?" or "why?". They lack imagination.

3. Animals do not have the higher emotions that require a conceptual repertoire

such as regret, gratitude, shame, pride, guilt, etc.

4. Animals do not form complex relationships with others

Because it requires the higher emotions like gratitude and resentment, and concepts such as rights and responsibilities.

5. Animals do not get art or music

We can pay disinterested attention to a work of art (or nature) for its own sake, taking pleasure from the exercise of our rational faculties thereof.

6. Animals do not laugh

I do not know if the science/philosophy of laughter is settled, but it appears to me to be some kind of phenomenon that depends on civil society.

7. Animals lack language

in the full sense of being able to engage in reason-giving dialogue with others, justifying your actions and explaining your intentions.

Scruton believed that all of the above arise together.

I know this is perhaps a little OT, but I seldom if ever see these issues mentioned in discussions about AGI. Maybe less applicable to super-intelligence, but certainly applicable to the "artificial human" part of the equation.

[1] Philosophy: Principles and Problems. Roger Scruton

Re: AGI is an engineering problem, not a model training problem

#136

Post by taormina »

dcre wrote: post #40

The first premise of the argument is that LLMs are plateauing in capability and this is obvious from using them. It is not obvious to me.

Just ancedata, but they keep releasing new versions and it keeps not being better. What would you describe this as if not plateauing? Worsening?

Re: AGI is an engineering problem, not a model training problem

#137

Post by hyfgfh »

Svgs, date management, Http, so many simpler things we dont have solve and somehow people believe they will do it by putting enough money in LLMs when it cant count

Why some people understood when they tried it with blockchain, nfts, web3, AR, ... any good engineer should know principle of energy efficiency instead of having faith in the Infinite monkey theorem

Re: AGI is an engineering problem, not a model training problem

#138

Post by Davidzheng »

AIPedant wrote: post #56
tshaddox wrote: post #21

Earlier quoted context omitted.

> We don't know if AGI is even possible outside of a biological construct yet. This is key. A discovery that AGI is impossible in principle to implement in an electronic computer would require a major fundamental discovery in physics that answers the question “what is the brain doing in order to implement general intelligence?”

It is vacuously true that a Turing machine can implement human intelligence: simply solve the Schrödinger equation for every atom in the human body and local environment. Obviously this is cost-prohibitive and we don’t have even 0.1% of the data required to make the simulation. Maybe we could simulate every single neuron instead, but again it’ll take many decades to gather the data in living human brains, and it woul…

i'd argue LLMs and deep learning are much more on the intelligence from complexity side than the nice symbolic solution side of things. Probably the human neuron, though intrinsically very complex, has nice low loss abstractions to small circuits. But on the higher levels, we don't build artificial neural networks by writing the programs ourselves.

Re: AGI is an engineering problem, not a model training problem

#139

Post by missingrib »

AIPedant wrote: post #56
tshaddox wrote: post #21

Earlier quoted context omitted.

> We don't know if AGI is even possible outside of a biological construct yet. This is key. A discovery that AGI is impossible in principle to implement in an electronic computer would require a major fundamental discovery in physics that answers the question “what is the brain doing in order to implement general intelligence?”

It is vacuously true that a Turing machine can implement human intelligence: simply solve the Schrödinger equation for every atom in the human body and local environment. Obviously this is cost-prohibitive and we don’t have even 0.1% of the data required to make the simulation. Maybe we could simulate every single neuron instead, but again it’ll take many decades to gather the data in living human brains, and it woul…

That is only true if consciousness is physical and the result of some physics going on in the human brain. We have no idea if that's true.

Re: AGI is an engineering problem, not a model training problem

#140

Post by manoDev »

Right now, we just did the equivalent of a tech demo of Broca's area.

AGI would take making at least one full brain, and then putting many of those working together, efficiently.

I don't believe we can engineer our way out of that before explaining how the f. the wetware works first.

Post reply on HN