What will be true about safe AI?
3
250Ṁ645
2050
78%
It's mechanistically interpretable
74%
It is aligned by a constitution
65%
Nobody owns it
65%
Sovereign individuals can earn a living by aligning it
65%
It is fairly steered by us competing to align it
52%
It's not machine learning based
50%
It is symbolic GOFAI
50%
The constitution that aligns it is kept aligned by the game theoretic principles of a free and open decentralized market.
34%
It's source code exists as a market
10%
It is Krantz

When all the alignment researchers agree that we've achieved "provably safe AI" that demonstrably poses no existential risk, what properties will that AI have?

Similar in spirit to this prediction:

https://manifold.markets/Krantz/what-will-be-true-about-the-person-R9tEgLhOOp?r=S3JhbnR6

Will resolve when Eliezer Yudkowsky agrees we've achieved "provably safe AI".

Get
Ṁ1,000
to start trading!
© Manifold Markets, Inc.TermsPrivacy