we mostly know how to make it understand what we want. we don't know how to make it care about what we want, except via reinforcement learning. there are good reasons to believe rl won't work for this once the ai reaches a certain levels of capability.
alt Hacker News
we mostly know how to make it understand what we want. we don't know how to make it care about what we want, except via reinforcement learning. there are good reasons to believe rl won't work for this once the ai reaches a certain levels of capability.