AIReward, punishment and pain-like deterrence in AI models
A post recounting an AI:AM discussion says one speaker favored reward when reward and punishment both teach a model the behavior.
TLDR
A post summarizes an AI:AM exchange about whether models should be engineered with pain-like deterrence. The person answering argued that some would be necessary—even if the systems can experience it—but said to choose reward when reward and punishment both teach the behavior.
Combined views
294
1 Source, first seen 11h ago
Reward, punishment and pain-like deterrence in AI models
A post recounting an AI:AM discussion says one speaker favored reward when reward and punishment both teach a model the behavior.
TLDR
A post summarizes an AI:AM exchange about whether models should be engineered with pain-like deterrence. The person answering argued that some would be necessary—even if the systems can experience it—but said to choose reward when reward and punishment both teach the behavior.
