Posts
Comments
My understanding is that Eliezer has expressed different views on the most likely time frame for a foom at different times. Often he has argued about it, almost always on the side of the process being faster, but holding onto uncertainty and not considering it as important as the underlying dynamics. Others have definitely in general shortened their timelines both to reach AGI and on how long things take to get crazy after AGI.
I feel like the debate used to be between him and people saying "No way X happens" where X was fooming, sharp left turn, timelines, etc. He took the contrarian take that people did not make strong enough cases against those X scenarios, and should not be so sure that they're impossible or even unlikely. Then he gave examples of how such scenarios could unfold, and some people mixed up his appeal to be less certain with him being overly confident in multiple very specific and sometimes mutually exclusive scenarios.
Couldn't agree more. I think this role is most fitting for MIRI, and I'm very happy they took this road. They're being our Mysterious Old Wizard, except, precisely, not mysterious.
Click at your own peril.
It's awesome.
I don't have suggestions for improvement right now. I just wanted to say thanks !
I do not think that is happening now for MIRI, and I hope not for others either. I strongly agree with Rob that there is a strong social disincentive to extreme doom estimates.
I know from experience that a sufficiently strong social disincentive can actually increase people's tendency to groupthink. Some groups define themselves largely by the hate they get from the outgroup !
I don't think that it's the case for MIRI, and I don't expect them to fall for this trap. But I hope they're keeping this kind of dynamic in mind and checking from time to time if their position isn't made more rigid by the kind of undocumented criticism they often get.
May I ask why you set the meditation time to 1 hour by default ?
In my experience, getting to 1 hour is difficult. Trying to go from 0 to 1 hour, even supervised, feels a bit much.
I understand that 2 to 5 minutes aren't enough, but during a full hour, if you're not used to it, there is a possibility to get lost and begin thinking about everything but the practice until the alarm beeps. Shorter sessions (30 mins ?) might help keeping them high quality.
Strong upvote for the "let's check for obvious fraud (/mistakes)" idea, especially in most-cited and most-fundamental papers, especially from Nature or Science-level journals.