I don't quite get why this is seen as stalling or hitting a plateau rather than the point where research has to move on to the next step for AI development (assuming that's the goal).

LLMs were never going to be artificial intelligence. At best LLMs would be an analog for the language center of a brain, and it sure seems like they nailed it.

What they would need next is the other functions of a brain beyond just language - imagine a human if the only part of their brain that worked was the language center. OpenAI kind of gets to that with o1, but at least what I know of it the attempt seems like a pretty hacky proof of concept.

If they really want AI they are very likely going to have to make breakthroughs in algorithm and architecture to best serve other brain functions, similar to what LLM architecture did for language.

> At best LLMs would be an analog for the language center of a brain, and it sure seems like they nailed it.

Unless a LLM create a neologism 'en passant', without prompting, they did not.

Why is "without prompting" a requirement? If a hypothetical LLM comes up with the best neologisms that one can imagine, but only after prompting, then they're disqualified?

I think the question would be how specific the prompt would be in that scenario. Is the prompt having to effectively describe the new idea, or is the prompt describing a general problem space and goal while letting the LLM do all the heavy lifting of innovation?

Either way, my issue there is actually with the fact that the LLM can only predict what a human might say in response to the prompt andz as far as we know, don't have any way of logically thinking through the new solution or modelling reality to understand if the idea might actually work.

It's not a requirement to be useful tools, its a requirement to be an "analog for the language center of a brain". I personally use them as boilerplate-generator and as rubber ducks, but a neurotypical brain will create neologisms quite often (accross a life, not everyday ofc) when talking to people.

I never personally expected anything close to neologism out of LLMs, not sure whether that was a common expectation.

LLMs would have to completely go outside the bounds of how they were designed to work for that to happen. They're prediction engines. They could absolutely create a novel idea when predicting what a person might say, but LLMs have no way of working through the logic behind the idea or modelling reality to see if the idea might work. You could task an LLM to design a thousand novel gene therapies or something similar, but at the end of the day humans would have to sift through testing or logically invalidating each option generated. Until that step is automated and paired with an LLM, and almost certainly other structures along the way, it will never be artificial intelligence.

> I never personally expected anything close to neologism out of LLMs, not sure whether that was a common expectation.

You just described how LLMs have not "nailed it" at being an "analog for the language center of a brain." Creating neologisms is something our language centers do.

> They could absolutely create a novel idea when predicting what a person might say, but LLMs have no way of working through the logic behind the idea or modelling reality to see if the idea might work.... but at the end of the day humans would have to sift through testing or logically invalidating each option generated.

That's not too different from a monkey at a typewriter.

It's so obvious, but try telling a tech bro that AI doesn't exist and they will have a cognitive dissonance meltdown. They fell for the propaganda and memes.

"A sufficiently-advanced technology is indistinguishable from magic."

That probably depends a lot on which group of tech bros you're talking about there. I'd expect most people working on LLMs every day understand the limitations and know that "AI" as a term was bastardized for marketing and fund raising.

wow this dropped off the front page in the space of a blink

https://hnrankings.info/42124409/

I suppose this sort of news could adversely affect certain VCs AI dominated portfolios

https://www.ycombinator.com/companies?batch=F24&batch=S24

this drop seems very odd. why is this happening?

It's the hardware I think. the plasticity of your wetware is what could be the cause of your thinking ability.

Buy SOXS

IMO:

Industrious and skilled people but _they_ are _the_ limiting factor, now, whatever methods they are going to come up with.

Remove yourselves from these projects and start from scratch.

nobody's stopping you from starting from scratch?

You are, of course, completely right.

So let's freeze frame right here. What is this thing actually good for? Give me even one killer app worth one hundred billion dollars.

Automate creation of the content which must be created but is useless and ignored by everyone. Like advertisements, both short and long. And then automate receiving and deleting such useless content. It's like a self-propagating smam+antispam machine. Same with pictures and maybe videos soon. Some pictures much be placed in some spots, like headers (aka ads) of the articles, and are now massively replaced with generated content. After we will be swamped with these pics, people will start working on filtering them with the neural nets. Very useful tech as we can see :)

> So let's freeze frame right here. What is this thing actually good for? Give me even one killer app worth one hundred billion dollars.

Propaganda bots to flood social media.

And most other forms of advertising. Anything where objective truth doesn't matter and quantity is more important than absolute quality is a great fit for LLMs.

This reminds me of the television with Philo Farnsworth and RCA. Just because OpenAI came up with the current mehthod and has a lot of resources, doesn’t mean they will ever invent it as some genius inventor will probably come up with the solution and may or may not work for OpenAI.