[resfirestar]

Implicit in a lot of "AI jobs apocalypse" predictions is the assumption that most tasks are ridiculously easy compared to AI research, so naturally the smart AI researchers can understand any profession well enough to credibly predict that AI will be able to replace it. I'm personally not sure the apocalypse has been truly disproven as opposed to progress just being slower than some of the overexuberant predictions, but there does seem to be a pattern of famous AI researchers predicting a job would be automated and turning out to be wrong because they focused too hard on a single aspect of it that could be automated while handwaving or ignoring the hard parts. This has prominently happened with radiology, then with customer service, and now they are walking back on programming too. Maybe take these guys with a grain of salt going forward? I trust them to be able to tell us frontier AI models will keep getting better, not to predict the impact that will have on specific industries. Some people will insist we should give them half credit for predicting there would be impact at all (as opposed to the "it's a bubble" refrain) but I think it should be possible to ignore two categories of obviously dumb predictions at the same time.

[smoe]

I think we too often treat other people’s jobs like spherical cows out of ignorance. Not just AI researchers.

Long before LLMs, programmers regularly and massively underestimated how hard it is to automate other people’s work. Knowledge workers often think carpenters just bang nails into wood, while blue collar workers think knowledge work as sitting in front of a screen copying values from Excel on the left into a form on the right while sipping a latte.

Only like 2.5 years ago, I thought programming would be one of the last knowledge worker jobs to be significantly affected by LLMs, not one of the first. I think AI models will continue to be very impactful. But for quite a while, they may mostly turn knowledge work into a last mile problem rather than eliminating it.

[zozbot234]

Programming has been successfully automated though. Programmers used to write programs line-by-line in raw binary code or assembly mnemonics, now they just write high-level formal code in languages like C++ or Rust and the computer spends much of its working time processing those lexer and parser 'tokens' and translating the whole thing into assembly and binary code. It all works quite well.

[gyomu]

Before:

- programmers spend time in meetings discussing requirements

- programmers spend time thinking how to improve performance and reliability

- programmers spend time tracking down issues in existing code

- programmers write binary/assembler code

Now:

- programmers spend time in meetings discussing requirements

- programmers spend time thinking how to improve performance and reliability

- programmers spend time tracking down issues in existing code

- programmers write C++/Rust code

Pray tell, where do you see the “programming has been successfully automated” part?

[rdedev]

There is a wide chasm between writing code in python vs "write a star craft clone". And that is not where near writing python vs writing binary code.

To put in another way, we have been building abstractions to make things easier for us to code. With coding agents you don't even code in the first place. It almost feels like a logical fallacy to compare the two