Will Luria replicate?
Jul 2, 2032
M$61 bet
Scott recently wrote about Luria asking questions to some Uzbek peasants in the 1930s here: https://astralcodexten.substack.com/p/somewhat-contra-marcus-on-ai-scaling?s=r The thing is, Kyrgyz and Uzbeks (after 80 years of mass literacy) are not exactly high in the IQ department TODAY -Kyrgyz scored around the equivalent of 75 on the 2009 PISA. Given it would be trivial for someone to visit Central Asia today (Kyrgyzstan would be easiest for a number of reasons), would Luria's observations replicate today? Resolves if someone travels to Uzbekistan/Kyrgyzstan/Tajikistan/southern Kazakhstan with this intention after I've posted this question and writes a detailed account of his observations. If the observations do not replicate Luria's at all, this question will resolve No. If they strongly replicate, it will resolve Yes. If it's a mixture, it will resolve to % on the basis of the agreement with Luria's observations.
Michael Wheatley is betting NO at 35%
Seem like the base rate for sufficiently interesting findings to replicate is low enough as it is.
As for the AI discussion in the comments of the SSC post, I find most of it funny. It's primarily focusing on irrelevancies while ignoring actually useful stuff. Thus my market here (actually useful): https://manifold.markets/EnopoletusHarding/will-the-discord-explicit-content-f I have two tests for AGI: if it can pass the Turing test and if it can identify any object presented to it. Whether these can be answered by scale is an empirical question.