Hmm. What is an example of a 'real craft'? I'm not sure I follow.
E.g. in no particular order: metal smith (silver gold etc), welder, carpentry, plumber, mechanic, fine artist, sculptor, textile. Basically the 'real trades' that have been around and should be around for many years.
The type of jobs that the white collars can't do without.
Either you outperform other CEOs and get rich on your own business, or you go this other path as I exemplified :D.
Now the gray zone is your regular 9-5, probably good co-workers, you make good living and work to live. The downside is that flexibility. So you need to think about what trades are you willing to make. I didn't mean to sound so 'dramatic' and pretend it's all black and white.
Industrial production has had a huge impact on many of those and will continue doing so. It‘s wrong to romanticize these jobs, they are just a likely to be automated.
I think it's much more likely LLMs become good enough to replace programmers before we have robots replacing plumbers.
That would be utopia.
Might as well say to get a real job like hunting, foraging for berries or animal herding.
Granted it might be still cheaper for humans to do the labor for some time, but eventually we will have most houses manufactured (home printing) or done as a prefab and assembled. For luxury some could be custom made by AI powered robots, based on user prompt.
I don't think that AI can reach such a level to install and measure everything without a carpenter. The gap is quite big and requires data which doesn't exist in the real world to feed into a model that can power AI to such an exact degree.
Look at how far we've come with LLM, that's only because we have so much texts readily available to train those models. In real life you'd have to manage a dataset combining different configurations of installations, even within prefab and assemblies you cannot reach the accuracy in my opinion without carrying a huge risk.