Major caveat at the bottom: “The researchers admit that these estimates are rough and likely to be wrong.”
I’m curious how they define “computer programmer”, a job which has apparently even odds of being mechanized - especially when systems analysts/DBAs/software engineers have a < 5% chance.
There’s a kind of weird “no true scotsman” fallacy going on I suspect, when you look at the history of computer programming. We just define the kind of programming that can be automated as not “true programming”.
Assembling code? Nah, the assembler does that, so it isn’t really programming. Unrolling loops? That’s a compiler option. Laying out forms? Nah, the forms library does that for you. Turning UML models into JSON-RPC APIs? Nah, there’s a library for that …
I thought it was odd that the ability “to squeeze into small spaces” was significant, but there’s no mention of the ability to solve CAPTCHA puzzles or play golf with clients :-).