quetzal_rainbow

[Question] How do you shut down an es­caped model?

Train­ing of su­per­in­tel­li­gence is se­cretly adversarial

There is no sharp bound­ary be­tween de­on­tol­ogy and consequentialism

Where Does Ad­ver­sar­ial Pres­sure Come From?

Pre­dictable Defect-Co­op­er­ate?

They are made of re­peat­ing patterns

[Question] How to model un­cer­tainty about prefer­ences?

[Question] What liter­a­ture on the neu­ro­science of de­ci­sion mak­ing can you recom­mend?

[Question] What spe­cific thing would you do with AI Align­ment Re­search As­sis­tant GPT?

[Question] Are there any tools to con­vert LW se­quences to PDF or any other file for­mat?

quet­zal_rain­bow’s Shortform