I strongly doubt there'll be any substantive advances till they know what they're doing (see Rob Slade's comment, quoted above), and I strongly doubt that they'll have time to get that right. As it is, the offending firms are all running at a heavy loss. Once someone needs to pay some real bills in genuine money, end-user bills will go up^†, the market will relax back to normal, and LLMs will go back to simple pattern-matching tasks where they started (and where they do provably and reliably work).

Don't make the category error of thinking that what LLMs do is thinking: all they've done is win at the Imitation Game. I read recenty that LLMs simulate the memory part of the brain, and that the reasoning part of the brain is entirely elsewhere. Someone needs to go back to first principles.

* A business lack-of-strategy pioneered by Netscape.

† Orders of magnitude are being suggested.