There are no new ideas in AI, only new datasets

But producing something useful is a totally different thing from producing resilience in physical reality. That takes a world model, and I guess my suspicion is that an entity can’t build a world model without a long history of surviving in that world.

Put another way, you can never replicate what it’s like to burn your hand on the fire using only words. You could have a million people tell a child about what fire is like, the dangers of it, the power of it, the pain of it. But they will never develop an innate understanding of it that helps them navigate the real world.

Until they stick their hand in the fire. Then they know.

See here for example:

[1] https://arxiv.org/pdf/2410.18072

[2] https://arxiv.org/pdf/2411.02914v1

[3] https://openai.com/index/video-generation-models-as-world-si...

But even if you knew nothing about this topic, the observation that you simply couldn't store the necessary amount of video data in a model such that it could simply regurgitate it should give you a big clue as to what is happening.

Maybe? Are you just outlining the thesis, or saying this should be self-evident?

> It is an encoding of patterns in the world where the only thing that matters is if anybody or anything wielding it can map the encodings to and from the patterns they encode for (which is more of a sociological/synchronisation challenge).

Yes, and my point is, current genAI utterly fails in unpredictable / bizarre ways because it only mimics the abstract encodings, ignorant of patterns in the world. Obviously some people argue "next token prediction is all you need," but that's a claim that is far from self-evident.

> There are many parts of non-programming languages that similarly have nothing to do with embodiment. Some of them are even about incredibly abstract things impossible in our universe. One could argue that for many fields genius lies in being able to mentally model what is so foreign to the intuition our embodiment has imbued us with or to be able to find a mapping to facilitate that intuition.

I would say this misses the point. The meaningfulness of abstractions, even ones that are unintuitive or unphysical or illogical, come from our embodied experience. Our enjoyment of even the most absurd fiction comes from our ability to simultaneously comprehend what it is "about" and what is "possible." Both relate to experience-of-reality and mean nothing in a vacuum.

> Said otherwise: the experience our embodiment has given us might limit how well we can understand the world (Quantum Mechanics anyone?).

I agree that we are limited in some ways by our particular embodiment. e.g. There's a huge spectrum of sensory experiences - colors, sounds, smells ... - which we know other animals have that we do not.

As I understand, where we disagree is on the why. I would say our capacity for understanding comes from our embodiment, therefore it's only natural that the limits of our embodiment also limit our understanding. After all we could imagine that if we had direct sensory experience of quantum effects, we would understand QM better or at least easier. In some fuzzy way, (no embodiment => poor understanding) and (embodiment => better understanding) is evidence for (embodiment understanding). I suppose your argument is a counterfactual that we might be able to imagine a (no embodiment & some understanding) so (no embodiment =/> poor understanding), but I don't see the evidence that this is not just imaginable but actually possible in reality.