Difference between revisions of "Pascal's mugging and AI safety"
| Line 9: | Line 9: | ||
Notably, [[Eliezer Yudkowsky]] has consistently argued against paying up in Pascal's mugging. | Notably, [[Eliezer Yudkowsky]] has consistently argued against paying up in Pascal's mugging. | ||
| − | e.g. <ref>https://forum.effectivealtruism.org/posts/zjbxdJbTTmTvrWAX9/tiny-probabilities-of-vast-utilities-concluding-arguments#The__claimed__probabilities_aren_t_that_small</ref> | + | e.g. <ref>https://forum.effectivealtruism.org/posts/zjbxdJbTTmTvrWAX9/tiny-probabilities-of-vast-utilities-concluding-arguments#The__claimed__probabilities_aren_t_that_small</ref> <ref>https://slatestarcodex.com/2015/08/12/stop-adding-zeroes/</ref> |
==References== | ==References== | ||
Latest revision as of 22:16, 17 November 2020
Existential risk reduction via work on AI safety has occasionally been compared to Pascal's mugging. The critic of AI safety argues that working on AI safety has a very small probability of a very big payoff, which sounds suspicious.
The standard resolution seems to be:
- Point out that there are different levels of what "very small probability" means. Some people think 1% is very small, whereas in Pascal's mugging we are dealing with astronomically small probabilities such as 1/3^^^3.
- Argue that for probabilities like 1%, standard expected value calculations work fine.
- Argue that reducing x-risk from AI safety is more like a 1% chance than like an astronomically small chance.
Notably, Eliezer Yudkowsky has consistently argued against paying up in Pascal's mugging.
