Re:Adversarial Policies Beat Superhuman Go AIs
by voidfunc on Tuesday December 24, 2024 @10:24AM
(#42502466)
I don't think I see them as a win, but they're easily dealt with. AI will need analysts at the latter stage to evaluate the outputs but that will be a relatively short-lived problem.
[ Parent
| on HN ]
Re:Adversarial Policies Beat Superhuman Go AIs
by bubaumba on Tuesday December 24, 2024 @10:35AM
(#42502548)
> I don't think I see them as a win
Unavoidable, probably
> but they're easily dealt with. AI will need analysts at the latter stage to evaluate the outputs but that will be a relatively short-lived problem
That solves only to some degree. Hallucinations may happen at this stage too. Then either correct answer can get rejected or false pass through.