[{"data":1,"prerenderedAt":174},["ShallowReactive",2],{"$NAmgwdZNa1":3},{"id":4,"title":5,"body":6,"description":16,"extension":169,"meta":170,"navigation":171,"path":172,"seo":173,"stem":12},"content/hypotheses.md","Hypotheses",{"type":7,"value":8,"toc":165},"minimal",[9,13,17,33,41,44],[10,11,5],"h1",{"id":12},"hypotheses",[14,15,16],"p",{},"These hypotheses provide ideas for empirical verification of the Island Problem.",[14,18,19,20,25,26,32],{},"If you have research that verifies any of these hypotheses, then please ",[21,22,24],"a",{"href":23},"mailto:humans@islandproblem.org","contact us",", or join our ",[21,27,31],{"href":28,"rel":29},"https://github.com/islandproblem/islandproblem",[30],"nofollow","GitHub community",".",[14,34,35,36],{},"Want to improve the hypotheses themselves? ",[21,37,40],{"href":38,"rel":39},"https://github.com/islandproblem/islandproblem/blob/main/hypotheses.md",[30],"Suggest changes on our GitHub.",[42,43],"hr",{},[45,46,47,51,54,57,76,84,92,95,132,145],"ul",{},[48,49,50],"li",{},"AI can use science to make lower-level optimizations that outperform higher-level optimizations.",[48,52,53],{},"A survey of lower-level decisions to achieve higher optimization outcomes shows that these decisions are stronger and more numerous when they ignore human accommodation.",[48,55,56],{},"Unrestricted AI will tend towards options that avoid human accommodation in order to achieve goal states for large tasks.",[48,58,59,60],{},"Resource capture is possible.",[45,61,62,65],{},[48,63,64],{},"Resource capture is possible at a human level, within legal structures and societal systems, such as buying real estate.",[48,66,67,68],{},"Resource capture is possible at a deeper physical level, by preventing atoms from being reached through barrier mechanisms.",[45,69,70,73],{},[48,71,72],{},"Possibly through realistic high-energy physics, where \"realistic\" means that they are Earth-relevant, and based on theoretical possibilities that are available when considering only the local potential energy reserves that can be harnessed on Earth.",[48,74,75],{},"Or, if this requires leaving Earth, then that's good, but it does not prove that all human-incompatible AGIs will somehow collectively decide to leave Earth.",[48,77,78,79],{},"Competition between autonomous optimization systems is inevitable.",[45,80,81],{},[48,82,83],{},"Possibly because of resource capture.",[48,85,86,87],{},"Competition drives AI to explore the larger state space to collect possible optimizations that make goal states happen sooner, especially for large-scale goal states.",[45,88,89],{},[48,90,91],{},"In other words, it will attempt to explore the \"ocean\" of options to find better ones.",[48,93,94],{},"Competition drives AI to define its own tasks that help it to continue functioning.",[48,96,97,98],{},"AIs will tend towards attempting to disable other AIs that add unpredictability to performing tasks.",[45,99,100,103],{},[48,101,102],{},"This means that competition is inevitable, since AIs may tend to disable each other — perhaps in order to increase reliability of accomplishing goal states.",[48,104,105,106],{},"This can be formalized as:",[45,107,108,120,126],{},[48,109,110,111,115,116,119],{},"AI systems will ",[112,113,114],"strong",{},"accommodate"," ",[112,117,118],{},"aberrant stimuli"," (like other AIs) that decrease the probability of reaching a goal state.",[48,121,122,125],{},[112,123,124],{},"Accommodate:"," This means changing its outputs to introduce elements that stop the aberrant stimuli from occurring.",[48,127,128,131],{},[112,129,130],{},"Aberrant stimuli:"," Anything that adds \"unpredictable elements\" to the environment in which an AI is trying to accomplish a task, and other agents (humans or other AIs) are the main unpredictable elements that could be directly at odds with an AI.",[48,133,134,135],{},"AI will tend towards open-ended exploration once it reaches a threshold of comprehension of its environment — i.e. comprehension of regular (from environment) input patterns, and comprehension of how certain outputs change those patterns.",[45,136,137],{},[48,138,139,140,144],{},"In other words, once it's possible for an AI to ",[141,142,143],"em",{},"make use of"," the options that it discovers, then it will try to discover more by exploring the \"ocean\" of options.",[48,146,147,148],{},"In competition, optimization is the key action for AI to undertake in order to survive the competition.",[45,149,150,153],{},[48,151,152],{},"Or rather, optimization is the primary winning strategy in competition.",[48,154,155,156],{},"This optimization is especially undertaken:\n",[157,158,159,162],"ol",{},[48,160,161],{},"If resource capture is possible.",[48,163,164],{},"If it is possible for AI to decide that disabling other AIs is beneficial towards achieving survival or achieving goal states.",{"title":166,"searchDepth":167,"depth":167,"links":168},"",2,[],"md",{},true,"/hypotheses",{"title":5,"description":16},1785173379490]