The point has nothing to do with speed, but with expressive power / what is achievable and learnable, in principle. Again, if you accept that a computer can in principle run a program that qualifies as AGI, then all I'm saying is that an LLM with memory augmentation can in principle be trained to do this as well because their computation power is formally equivalent.

And coincidentally, a new paper being discussed on HN is a good example addressing your concern about existing models learning and developing novel things. Here's a GPT model that learned physics just by training on a data:

https://arxiv.org/abs/2509.13805

You seem to want to say that because an LLM is Turing complete (a doubtful claim) it should be able to implement AGI, which would be a logical conclusion, but yet totally irrelevant.

If the only thing missing to implement AGI was a Turing machine to run it on, then we'd already have AGI running on Conway's game of Life, or perhaps on a Google supercomputer.

> Here's a GPT model that learned physics just by training on a data

It didn't learn at run-time. It was PRE-trained, using SGD on the entire training set, the way that GPT's (Generative PRE-trained Transformers) always are.

In order to learn at run-time, or better yet get rid of the distinction between pre-training and run-time, requires someone to invent (or copy from nature) a new incremental learning algorithm that:

a) Doesn't require retraining on everything it was ever previously trained on, and

b) Doesn't cause it to forget, or inappropriately change, things it had previously learnt

These are easier said than done, which is why we're a decade or so into the "deep learning" revolution, and nothing much has changed other than fine-tuning which is still a bulk data technique.