OpenAI and others seek new path to smarter AI as current methods hit limitations | Hacker News Reader

OpenAI and others seek new path to smarter AI as current methods hit limitations

reuters.com

12 pointsisaacfrond9 comments

The process here seems relatively clear.

Original logic-based "AI" system didn't "fail-fail". They could reproduce some complex human behaviors but they required expensive people writing their code.

Deep learning succeeded because it allowed compute to be substituted for these coders and so companies could produce the current structures without an army of zillions of experts. But these systems still suffer from problems of "not really learning" or "not really understanding".

I don't think the problem of discovering how humans quickly learn to make effective novel responses to new information is unsolvable but the deep learning arc may not have brought it's solution much closer (despite appearances).

An LLM is like an idiot savant --- i.e. kinda like in the movie Rain Man.

They excel at remembering data but they often fail in applying all that data to reach any logical conclusion.

In other words, they are sort of a big, expensive, power hungry, talking hard drive --- which is a technical solution looking for a problem in a lot of cases.

The thing with AI isnt that it doesn't have a whole lot of use cases,it is thr fact that its use cases dont justify the hype or investment it has gotten.

It is somewhere in between Crypto and dot.com as far as bubbles go, and as open models catch up with closed source ones, it seems less and less likely that yhese companies are ever gonna make back the monry they spent, at least short term.

The thing with AI isnt that it doesn't have a whole lot of use cases,it is thr fact that its use cases dont justify the hype or investment it has gotten.

The truly disheartening aspect is the fact that so many technical "leaders" have bought into the hype and are throwing huge sums of money down this rabbit hole. It's not like we haven't been through this sort of thing before.

And many seem convinced that these talking databases that can't do simple math are somehow a threat to society. This is an insult to humanity in more ways than one.

ChatGPT's 200 million weekly active users worldwide would disagree with you.

So it's a useful idiot savant

The limitation should be obvious --- the expectation that real "intelligence" will somehow "emerge" from a statistical process is unfounded.

It worked pretty well for many applications. Machine translations are almost on par with human translations. This is a multi-billion dollar industry.

Multi-billi9m industry sure, but it doesnt justify near trillions of spending and hype.

AI is slightly behond search engines/internet in terms of usefulness, and even that was bubble, and the sad thing is we already had a few AI wonters before simply due to the cost of research, if they end up losing money, or worse we end uo in a recession because of AI hype, research and funding are gonna dry up ridiculously fast, which wouod hurt not only AI but most of the tech industry short term.

Well at least the us has recovered enough to absorb that fall, i hope.

Reply on news.ycombinator.com