The Seven Voyages Of SteveThe breathless folks who have watched one person make a basic prototype of a piece of software authored with an LLM and then extrapolating that this is the future are just illustrating how much they don’t understand the problem. Bashing out modified template code (copy/paste/modify) isn’t *remotely* the hard part of making software. The hard part is understanding the problem and figuring out why a solution doesn’t work quite right. LLMs can’t do the former, and make the latter *much* worse
I guarantee this won’t stop these people using it anyway and creating all sorts of half-assed things that *sort of*work but have all sorts of edge case problems that no-one understands. These things will costs 10x as much to fix later as to do right in the first place, and *maybe* this will be kind of the same as their experience of hiring certain big name contract software companies (I’ve had to mop up after these companies before, but at least their terrible code might have comments).
It doesn’t make a lot of difference to me, I’m crossing the threshold into perceived obsolescence anyway (ageism is definitely a thing in this industry), but it saddens me that this will be yet another weapon in the arsenal of people who don’t really understand the difference between good and bad software to drag the average down even further. I’d like to think it will be used primarily as an advisory/auditing tool than a generative one, but I’ve been around too long to believe that
elFlashor@sinbad We have to brace ourselves for a new generation of publicly used software (in stores, public transport, banks, etc...) that have been cheaply made with LLMs.
They will work even worse than the current ones, with no money to fix them or even try to, and face even more excuses of the sort "sorry I cannot fix your (albeit very simple!) issue here, the software doesn't let me".
All that was already happening, but it's just gonna get worse and worse...