Hacker News Reader

Comment by famouswaffles | Hacker News Reader

It's a lot more than just commercially meaningful. Even if you don't care about money, "highly autonomous systems that outperform humans at most economically valuable work" has a lot more weight on "how close we are to achieving human intelligence" than most of the 'AGI' thresholds I've seen yet.

View on HN

Traditionally, AGI means being capable of learning everything (not necessarily at once) that can be learned by the same AI agent. It differs from the commercially meaningful definition in that a standard specific-purpose-built AI can still do most tasks if you give it enough specific purposes, but it will be just as incapable of human intelligence as ever.

No - the term "AGI" was really coined to distinguish general intelligence from narrow intelligence(s). There's an obscure earlier usage, but it basically became mainstream as the title of a book edited by Ben Gortzel who had in turn got it from Shane Legg.

Shane Legg would a few years later go on to co-found DeepMind, with creating AGI as their declared mission. Legg's personal definition of AGI is not just generalist AI, but specifically human-level generalist.

Well, let's learn to walk first.

Right now we have just massively jagged intelligence that performs stunning feats in math and single-shotting three.js games, and falls flat on it's face in many every day scenarios.

And, it can't even learn.

All intelligence is jagged. Human intelligence is jagged, so is Cetacean intelligence, so is Ape intelligence, so is Corvid intelligence. You could re-orient that statement for any species of intelligence relative to the other, or even individuals within a species. The big problem here is people being so convinced of their specialness they'll just keep sticking their heads in the sand, until we essentialy create God, and by then none of this useless pedantry will even matter. If you can't see we've already learnt to walk then there's nothing more to discuss here.

I'm not sure it's accurate to call human intelligence jagged - it seems (perhaps largely thanks to language) that we're able to make progress on anything we set our minds to, and of course we're now in process of building AI that will eventually be able to do things that we ourselves couldn't, whether due to computational demands, or memory capacity, or having sensory inputs that we don't have, etc. But at the end of the day, AI is a tool we've built, so maybe it will make us less jagged, and certainly more capable.

In any case, human intelligence as a goal is special since we live in a human world, and one of the prime goals of AI is to be able to do human jobs. If a human can do something and AI can't then that is a limitation, while if a human can't and AI can't either, that doesn't matter for the time being.

Yeah, I shouldn't have said AI can't walk yet, but most of what it can do is just due to scale - we're still just building big transformers, basically using the same 10 year old architecture that accidentally set us off on this path. I'm not sure we're really progressing towards human/animal type intelligence, just learning how much of our own capability can be realized by automating language.

I'm not sure it's meaningful to compare across different types of intelligence--but I don't think human intelligence is so special that we can pretend it's much less jagged than all other animals for sure. Our scales of models are also not far off biological, probably less efficient but not like astronomically so i would guess.

But anyhow I agree that inference-time learning doesn't exist and it's a big issue still I think.

It can learn very well within its context. And it can leave notes for itself to read into future contexts.

That's learning, whether you agree or not, and whether you like it or not. Baby steps, yes, but walking nevertheless.

There seems to be very few cases where ICL is really doing something that can be considered as learning (albeit ephemeral) rather than just utilizing in-context data via induction heads.

Yes, I'm aware of the linear regression example, and a few others, but these appear to just be specific capabilities that were learnt during pre-training, presumably pursuant to reducing errors on similar-but-different training samples, not any kind of generic run-time learning capability.

Reply on news.ycombinator.com