Three years ago I asked the same question (https://news.ycombinator.com/item?id=35327092), and the landscape has changed since then. Just say that even the most AI skepticism people will have to admit that with recent models, even open source ones, LLM is already a valuable intelligence resource, at least in the right hands. Of course we are nowhere near AGI yet, but no theorem can assert that AI will not be at least as good as human; neuro-networks can fit arbitrarily complex function. If human thought, based on the physical structure of brain, is computable, then some computer program can certainly take advantage of that. Even if there is some weird physics that yields non-computable properties, there is also no stop of building such a computer to take advantage of that law and thus have an AI on top of it too.

Some disciplines are invulnerable to AIs, such as math or art, where human comprehension and expression is the product itself. But how about others such as engineering where "if it works it works"? There are certainly engineers who love engineering but instead of "here is some AI tools that lets you focus on your main, non-trivial tasks", they are getting "you can certainly do that, but AI will complete faster and with less money". (A lot of people had trivial work that can already largely be replaced by AI by now.) AI might take a large portion of fun out of these fields. Maybe they will degrade to sports or specialized handcrafts, just like running (sports) or sewing (handcrafts) in modern world.

I wrote https://news.ycombinator.com/item?id=49578280#49579794 several days ago, but on second thought I can't even convince myself. We have to admit that sometimes we want to do something because it is hard, or never done before. External incentives are also totally valid options.