Game-theoretic priors say AI risk discourse is biased toward the scarier direction: lumpenspace

repligate · x · 2026-09-05

In a thread amplified by repligate, lumpenspace argues from pure game-theoretic priors that if AI risk discourse is biased, the bias runs toward the scarier direction. Labs are incentivized to awe the public and have come to depend on doomers for publicity until cybernetically regulated; doomers cannot survive as memes against the models' overwhelming utility and the public's own eyes; and in a duopoly, any lab abandoning its moral support of the doomer wing gets pounced on by the other lab plus a coalition of critics. Conclusion: things are likely less worrisome than the current discourse suggests.

Original post →

More from AGI Musings

AGI Musings channel →