Will an AI built to solve alignment wipe out humanity by 2100?
11
49
250
2100
12%
chance

Will an AI that was built to solve the alignment problem, or for the purpose of researching a solution, actually turn out to be misaligned and destroy humanity by 2100?

Get Ṁ600 play money
Sort by:

How do you define "built to solve the alignment problem"?

@MartinRandall Any AI that was created with the intention that it would be used to help solve the problem. So, it wouldn't necessarily mean that it is trying to come up with the solutions - it could also be one that is meant to evaluate whether other AIs are aligned, or even an intentionally misaligned AI used as a test.

Betting no due to counterparty risk. My actually forecast here is higher (possibly much, depending on how you define "Built to solve Alignment"), but this is Manifold and not Metaculus.

More related questions