- 7 weeks ago, I precommitted that censoring a post or comment on LessWrong would cause a 0.0001% increase in existential risk.
- Earlier today, Yudkowsky censored a post on less wrong
- 20 minutes later, existential risks increased 0.0001% (to the best of my estimation).
Wow, that's even more impressive than the claim made by some Christian theologians that part of the enjoyment in heaven is getting to watch the damned be tormented. If any AI thinks anything even close to this then we have failed Friendliness even more than if we made a simple object maximizer.
Next thing you're going to tell me that an FAI shouldn't push fat people in front of trolleys.
Note: A sufficiently powerful FAI shouldn't need to, but that is different from saying it wouldn't.