@conjurial 1 is full speed ahead to ASI, but inherently in the nature of superintelligece it is not xrisk-scale murderous or accidentally what we build isn't 2 is AGI -> ASI progress is slow enough (constrained by hardware / data?) that we "solve alignment" with safety tools
More notes on nearby ideas
Reply · Jan 25, 2023@personofswag or that AGI itself isn't xrisk and perhaps slow takeoff to ASI?Reply · Feb 22, 2023@jim_rutt @levitimg AGI is similar-to-human capabilities, vs ASI is significantly smarter than any humans... there's a lot of uncertainty about how we…Reply · Nov 7, 2023@RokoMijic sighhhh.... see this is where I get frustrated w Safety rhetoric- you want the AI to be aligned to do whatever the human wants them to do…Reply · Feb 22, 2023@TetraspaceWest and to disambiguate the first option, I'm thinking more of a situation where ASI doesn't lead to xrisk more because intrinsically it…