←back to thread

AI should only run as fast as we can catch up

(higashi.blog)
196 points yuedongze | 1 comments | 08 Dec 25 17:38 UTC | HN request time: 0.194s | source
Show context
gradus_ad ◴[08 Dec 25 17:51 UTC] No.46195373[source]▶
The proliferation of nondeterministically generated code is here to stay. Part of our response must be more dynamic, more comprehensive and more realistic workload simulation and testing frameworks.
replies(5): >>46195431 #>>46195733 #>>46197437 #>>46197956 #>>46199307 #
yuedongze ◴[08 Dec 25 17:56 UTC] No.46195431[source]▶
i've seen a lot of startups that use AI to QA human work. how about the idea of use humans to QA AI work? a lot of interesting things might follow
replies(6): >>46195474 #>>46195546 #>>46195718 #>>46195741 #>>46195828 #>>46199496 #
1. hn_acc1 ◴[09 Dec 25 00:03 UTC] No.46199496[source]▶
This feels a lot like the "humans must be ready at any time to take over from FSD" that Tesla is trying to push. With presumably similar results.

If it works 85% of the time, how soon do you catch that it is moving in the wrong direction? Are you having a standup every few minutes for it to review (edit) it's work with you? Are you reviewing hundreds of thousands of lines of code every day?

It feels a bit like pouring cement or molten steel really fast: at best, it works, and you get things done way faster. Get it just a bit wrong, and your work is all messed up, as well as a lot of collateral damage. But I guess if you haven't shipped yet, it's ok to start over? How many different respins can you keep in your head before it all blends?