AI safety researcher Joshua Achiam co-signed the critique.
One of the main reasons I criticize the AI safety community so much is because it’s the single biggest hub of people who might plausibly make large and valuable contributions to the world, if only they were better at updating from experience and criticism.
Users strongly agreed with critiques that the AI safety community fails to update its views from valid external criticism.
No Digg Deeper questions have been answered for this story yet.
Approximately co-signed. I wouldn't say "the" single biggest but certainly one of them.
One of the main reasons I criticize the AI safety community so much is because it’s the single biggest hub of people who might plausibly make large and valuable contributions to the world, if only they were better at updating from experience and criticism.
This seems worth stating explicitly to balance the negativity.
Unfortunately it *doesn’t* mean that the community is currently making small valuable contributions: near misses often involve causing the opposite of what you wanted. More on this here: https://www.mindthefuture.info/p/on-pessimization
(Though accelerating AI progress might still end up being accidentally good for the world!)
One of the main reasons I criticize the AI safety community so much is because it’s the single biggest hub of people who might plausibly make large and valuable contributions to the world, if only they were better at updating from experience and criticism.
Hmm so I should have said “large and robustly valuable” in the original tweet, to rule out the cases where pessimizing your own goals accidentally turns out to have good consequences.
This seems worth stating explicitly to balance the negativity.
Unfortunately it *doesn’t* mean that the community is currently making small valuable contributions: near misses often involve causing the opposite of what you wanted. More on this here: https://www.mindthefuture.info/p/on-pessimization
(Though accelerating AI progress might still end up being accidentally good for the world!)
@jachiam0 can you point to another? I think other communities could make bigger contributions given other changes, but the “better at updating” as a bottleneck (as opposed to human capital or lack of relevant knowledge) feels very distinctly AI safety.
Approximately co-signed. I wouldn't say "the" single biggest but certainly one of them.
@RichardMCNgo Heavy +1