15
Modern software quality, or why I think using language models for programming is a bad idea
(softwarecrisis.dev)
The current AI doesn't produce working software on the first try. It has to be guided into it, and still requires a decent programmer to end up with a decent product at the end.
Even if the AI got it correct the first time, the initial prompt is unlikely to be correct, for multiple reasons.
This all means that AI can merely assists decent programmers, it can't replace them. It can speed them up, even if only by taking a lot of the typing out of the equation. But it can't replace them.
AI could theoretically reduce the number of programmers in a company, but my experience has been that the company has a budget for programmers, and that's what they'll pay for. I've never cleared my to-do list at work, and I can no longer imagine it actually happening, no matter how much help the computer gives me.
I remain unconcerned about AI taking programming jobs.