Tim Bray is exactly right here. If Claude can impress Knuth, maybe people who don't know that should re-evaluate.
To me the danger in AI is who owns it. Like all powerful technologies, who owns and uses that power matters a lot. Nuclear is a great example. You can use it for power generation of WMDs.
Unfortunately a cabal of crazy people own LLM models, at least in the west.
Open Weight models are a force against this.