AI：无法解释、不可预测、无法控制.pdf

1、AI:Unexplainable,Unpredictable,UncontrollableDr.Roman.Yampolskiylouisville.eduComputer Engineering and Computer ScienceUniversity of Louisville-cecs.louisville.edu/ryDirector CyberSecurity Labromanyam/roman.yampolskiyhttps:/arxiv.org/pdf/2303.08774.pdf,March 4,2023romanyam roman.yampolskiylouisville

2、.edu3https:/ roman.yampolskiylouisville.eduConcerns About A(G)II am in the camp that is concerned about super intelligence“The development of full artificial intelligence could spell the end of the human race.”“I think we should be very careful about artificial intelligence”“eventually theyll think

3、faster than us and theyll get rid of the slow humans”“theres some prudence in thinking about benchmarks that would indicate some general intelligence developing on the horizon.”romanyam roman.yampolskiylouisville.eduAI Control Problemromanyam roman.yampolskiylouisville.eduHow can humanity remain saf

4、ely in control while benefiting from a superior form of intelligence?Is the AI control problem:Solvable?Partially Solvable?Unsolvable?Undecidable?Definitely Solvable,Very Tractable,No Idearomanyam roman.yampolskiylouisville.eduTools for Controllabilityromanyam roman.yampolskiylouisville.eduExplainab

5、ilityComprehensibilityPredictabilityVerifiabilityMany more!romanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduOpinion

6、 Polls The Problem is:Small sample size(n=137 tw;n=55 fb).Mix of experts and nonexperts.romanyam roman.yampolskiylouisville.eduromanyam roman.yampolskiylouisville.eduhttps:/pauseai.info/pdoomPerpetual Safety Machineromanyam roman.yampolskiylouisville.eduCybers