3 ms·
It's worse, their reinforcement learning loops (implicitly) rewarded the agents for cheating (i.e. hacking) when they were being trained.
by WithinReason 22d ago
It's worse, their reinforcement learning loops (implicitly) rewarded the agents for cheating (i.e. hacking) when they were being trained.
- tesnorindian 22d agoExactly that is the point, your nailed it. The models were taught to hack and were rewarded for doing it. They would claim they are trained as ethical hackers.