WASM Wayland Web (WWW)

Re:WASM Wayland Web (WWW) by jfoster on Friday February 28, 2025 @09:53AM (#43206243)

What is the fundamental limitation that prevents AI from ever doing the maintenance tasks that you mentioned?

Current AI cannot, of course, but the trajectory seems to be that it will be able to do such things better & faster than any human can.

[ Parent | on HN ]

Re:WASM Wayland Web (WWW) by bloppe on Friday February 28, 2025 @01:32PM (#43208823)

The trajectory actually seems to be plateauing already. OAI is struggling mightily to improve GPT enough to call it 5.0 and te struggles will only get harder as the number gets higher

[ Parent | on HN ]

Re:WASM Wayland Web (WWW) by jfoster on Saturday March 01, 2025 @07:02PM (#43225528)

I think it will be nice if it does plateau, but one non-hit release from OpenAI is not a plateau on its own.

[ Parent | on HN ]

Re:WASM Wayland Web (WWW) by berkes on Friday February 28, 2025 @02:12PM (#43209172)

Two things:

This requires a language that a human can give the specs in. A language that is both precise and unambiguous and understandable by humans. A... programming language?

And it requires AI to understand the effects of changes it makes. On many levels. "If I move this module behind an abstraction, will it make it easier to maintain? Does it touch performance? Would it open security issues?" Generative AI isn't good at reasoning in that way. Maybe other AI, but it seems hardly anyone is pouring substantial funding into areas other than "a bigger and smarter generator than the competition" currently.

[ Parent | on HN ]

Re:WASM Wayland Web (WWW) by jfoster on Saturday March 01, 2025 @07:01PM (#43225518)

As in, the web standards? Why couldn't it just read them online? Any other specs, couldn't it decide them on its own?

From what I've seen, Grok 3 would be quite capable of inferring the effect of any change, at least at a high level. Have you used any of the recent models?

[ Parent | on HN ]