They're making pretty good progress on hands[1,2], I don't have a good sense of how quickly these things will become cheap enough to become ubiquitous, I'd imagine > 5 years but < 10 years. One Policy and Zero Prior Learning are making very good progress[3,4,5].
Plummer is probably the safest of the trades if I had to guess. That said, I'd encourage you to think more about other things humans are likely to always want/need - see if any of the other options are interesting to you? For example, lots of people are and will continue to be lonely, having a bar to sit at with someone friendly/interesting is a tool we use for this (I wrote a blog post about this idea: https://b.h4x.zip/finale/) - but there are also things like undertakers, kids activities centers/camp councilor, etc. Not all of these things will be equally financially rewarding, but maybe you're able to think about rewarding in other ways. :)
[1]https://www.youtube.com/watch?v=4whgw2gLBS8
[2]https://www.youtube.com/watch?v=C5MsqzQW8X4
[3]https://www.youtube.com/watch?v=fhK28R9J-DQ
[4]https://nico-bohlinger.github.io/one_policy_to_run_them_all_...
They're nowhere near close with those blocky, slow, and confused hands. The ChatGPT moment for robotics will occur when you see a robot with the hand dexterity of a 6 year old.