4 ms·

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 (ass

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.

orwin 2y ago
> 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.
- uh_uh 2y ago
  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?
  - _heimdall 2y ago
    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.
  - orwin 2y ago
    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.
- _heimdall 2y ago
  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.
  - tivert 2y ago
    > 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.
nathanaldensr 2y ago
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."
- _heimdall 2y ago
  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.

orwin 2y ago

> 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.

uh_uh 2y ago
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?
- _heimdall 2y ago
  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.
- orwin 2y ago
  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.
_heimdall 2y ago
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.
- tivert 2y ago
  > 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.

nathanaldensr 2y ago

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."

_heimdall 2y ago
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.