There's practically no reason to think that AI will greatly accelerate the development of new algorithms.

Why? Because we're nearing the theoretical limits or because AI will be too dumb to help or for other reasons?

For the same reason that there's no reason to think AI will continue to advance exponentially.

The idea of the singularity is fun, but it's unrealistic. Nothing lasts exponentially forever.

Even if the exponential improvements, which have been happening in both hardware and AI algorithms, were to slow down to sub-exponential, this still assumes that exponential advancements are necessary for AI to reach a level where it can assist in algorithm design. And only now we're seeing big jumps in AI investments, which could serve as another source of improvements. You could even argue that we're seeing the first signs of recursive improvements, with Copilot making programmers ~20% more productive.

I have yet to see an AI bot produce anything that's truly intelligent and/or original (all I see is ever more powerful hardware and ever bigger quantities of training data being thrown at essentially the same statistical probability algorithms), and I don't predict that changing in the foreseeable future - at least, not without the same kind of fundamental breakthrough that would be required for quantum computing to become a practical reality.