I can't say everything, but with the current trend, Machine will plant, grow and harvest food. I can't say for open heart surgery because it may be regulated heavily.
Open heart surgery? All that's needed to destroy the entire medical profession is one peer reviewed article published in a notable journal comparing the outcomes of human and AI surgeons. If it turns out that AI surgeons offer better outcomes and less complications, not using this technology turns into criminal negligence. In a world where such a fact is known, letting human surgeons operate on people means you are needlessly harming or killing some of them.
You can even calculate the average number of people that can be operated on before harm occurs: number needed to harm (NNH). If NNH(AI) > NNH(humans), it becomes impossible to recommend that patients submit to surgery at the hands of human surgeons. It is that simple.
If we discover that AI surgeons harm one in every 1000 patients while human surgeons harm one in every 100 patients, human surgeons are done.
And the opposite holds, if the AI surgeon is worse (great for 80%, but sucks at the edge cases for example) then that's it. Build a better one, go through attempts at certification, but now with the burden that no one trusts you.
The assumption, and a common one by the look of this whole thread, that ChatGPT, Sora and the rest represent the beginning of an inevitable march towards AGI seems incredible baseless to me. It's only really possible to make the claim at all because we know so little about what AGI is, that we can project qualities we imagine it would have onto whatever we have now.
