Most of those researchers would disagree with your statement that the next major breakthrough is probably decades in the future.

Many of the insights accumulated in the decades up to now will probably prove useful for creating non-LLM AI, and the researchers can use LLM AI to speed up their research into non-LLM AI.