[nathan_compton]

I say that large language models are not intelligent because of the way they fail to do things. In particular, they fail in such a way as to indicate they have no mental model of the things they parrot. If you give them a simple, but very unusual, coding problem, they will confidently give you an incorrect solution even though they seem to understand programming when dealing with things similar to their training data.

An intelligent thing should easily generalize in these situations but LLMs fail to. I use GPT4 every day and I frequently encounter this kind of thing.

[NumberWangMan]

Is there a definition of intelligence that rules out large language models, but that does not also rule out large portions of humanity? A lot of people would readily admit that they don't have programming aptitude and would probably end up just memorizing things. Do we say those people are not intelligent?

It seems to me that the perceived difference is mostly in being able to admit that you don't know something, rather than make up an answer -- but making up an answer is still something that humans do sometimes.

[nathan_compton]

I have to admit this is a genuinely interesting question. Language models demonstrably do have some models of the world inside of them. And, I admit, what I say that they aren't intelligent, I mostly mean they are very stupid, rather than like a machine or algorithm. Artificial stupidity is progress.

[pixl97]

Ok, so from your other comment, I think this is where our definition of intelligence is breaking down...

Biological agents have a consistent world model based on their capabilities because an inconsistent model would lead to lack of reproduction or death. We could call this environmental intelligence.

Meanwhile we have LLMs that have appear to have what I would consider 'micro' world models for some things, but not a large consistent world model. I'm guessing this is due to a few things, but for example not being culled for bad world models would be one, and another is they are only grounded in text and we've not really explored multi-modal grounding in models very far.

I guess what's going to be interesting is to see how multi-modal and embodied models do as they are trained in the environment and create a more consistent world model.

[nathan_compton]

I believe that the best way to understand these large language models is that they have models of patterns of text. To the extent that patterns of text are congruent with patterns in the world, they appear to function well, but I think, in the end, they are statistical models of text, not of the world, and that substantially limits their capabilities.

I do think multi-modal models will be interesting, but text is a very special sort of thing. It is widely available, semantically rich, and informationally pretty dense. I'm not sure there is such a nice set of properties for other modes. Consider that we have already almost reached training data exhaustion with text and it is, by far, the most voluminous/dense training mode there is.