I would understand to read this kind of articles in 2023ish, but for what they can do now... It is not even if they are "intelligent" is what intelligence even is. I think it is undeniable that there is a level of capacity increasing with each new model, and how would you call that?
There is an illusion in LLM's, but that illusion, in my point of view, is about how much of an "entity" each chat bot is, not if they are capable of mimic intelligence or not.
LLMs are not intelligent for the simple reason that if you had as much knowledge as an LLM has, you would be churning out deductions and new discoveries every second. You'd be putting out Nobel-prize-worthy research every month.
LLMs are like a 50 IQ person with no intrinsic motivation but unbelievably vast knowledge and near-perfect recall.
I don't think intelligence needs to take the shape of human-like intelligence, nor human-like data efficiency.
I'd argue they are intelligent entities, superhuman in some ways due to being run on silicon but extremely data inefficient with nonobvious failure modes.
>I don't think intelligence needs to take the shape of human-like intelligence
But then the word doesn't mean anything. Intelligence is almost by definition what we're like.
If you think LLMs are intelligent, you have to concede that databases are also intelligent. And if you can't then you need to ponder deeply about what quality of an LLM's response to your query makes it (seem) more "intelligent" than a database's response to your SQL query, beyond the trivial thing that it's in natural language.
No, it's not. We don't really have a good definition of intelligence. If we can imagine non-carbon based life forms, I don't think its a leap to imagine non-human brain based intelligence (octopi?)
> If you think LLMs are intelligent, you have to concede that databases are also intelligent.
No, I don't. The mechanism by which LLMs answer and reason through problems is different, so I simply don't think these are comparable.
reply