The thing that I think all try to not talk about is:

What if the LLM capability plateaued (or about to)?

There is a limit to how much an LLM can probabilistically predict based on the training, and there is a limit to available knowledge, at least accessible knowledge.

So what happens if we hit the plateau in the next months?

Consider the incentives of AI companies in a situation like this. Consider the market’s reaction.

Scary times ahead.