Post #2769035
2026-03-18 14:41 UTC
@happyborg@fosstodon.org It's not entirely the same problem as earlier attempts at making natural-language compilers.
The "large" part of LLM gives it more context and a bit of "common sense". It's still a guess, but drawn from a more likely distribution.
The learned context is so strong they can sometimes surprise Babbage and give right answers to wrong questions.
Yeah, they routinely fail at software engineering we have today, but I wonder how much the world will adapt to "cheap but inaccurate" code.
Replies (1)
-
@happyborg@fosstodon.org 2026-03-18 15:04
@kornel@mastodon.social > I wonder how much the world will adapt to "cheap but inaccurate" code. Isn't that a contradiction, as in inadequate -> costly outcomes? 🤔 It is maybe a valid question, but to me uninteresting except in highlighting an awful possible outcome. The last thing the world needs is billions of "script kiddies". We are seeing this, but I think only scratching the surface. To me it's just a way of accelerating collapse. It is interesting to watch if you can detach yourself from that aspect.