The human brain operates at better levels of intelligence than the best LLMs, at 20 watts of power. We are a long way to that kind of efficiency, it will probably take both bespoke hardware and algorithmic improvements to catch up to nature.

I think the idea is fundamentally improved architectures. For example, transformer-based models were an incredible stepwise improvement. Self improvement would be a model discovering a stepwise improvement similar to the transformer. And presumably the improved models from that would be more likely to make further advances still.

Learning from training data is technically self-improvement but not the sort that is typically meant in this context.

Only if you think in terms of perceived raw intelligence, but self-update is a form of valuable self-improvement that could benefit current models a lot, if they could commit facts from context into their weights cheaply and reliably.

> AIs already improve themselves via training

Marginally. Model collapse is still a problem. Continuous learning is still a problem.

For AI to make a big leap we need a big break through.