AlexMennen(Alex Mennen)

What is cal­ibra­tion?

Search­ing for a model’s con­cepts by their shape – a the­o­ret­i­cal framework

Event [Berkeley]: Align­ment Col­lab­o­ra­tor Speed-Meeting

Why bet Kelly?

Aver­age prob­a­bil­ities, not log odds

Map­ping Out Alignment

AlexMen­nen’s Shortform

When wish­ful think­ing works

Safely and use­fully spec­tat­ing on AIs op­ti­miz­ing over toy worlds

Com­pu­ta­tional effi­ciency rea­sons not to model VNM-ra­tio­nal prefer­ence re­la­tions with util­ity functions

A com­ment on the IDA-AlphaGoZero metaphor; ca­pa­bil­ities ver­sus alignment

Log­i­cal un­cer­tainty and math­e­mat­i­cal uncertainty

Log­i­cal un­cer­tainty and Math­e­mat­i­cal uncertainty

More on the Lin­ear Utility Hy­poth­e­sis and the Lev­er­age Prior

Value learn­ing sub­prob­lem: learn­ing goals of sim­ple agents

Against the Lin­ear Utility Hy­poth­e­sis and the Lev­er­age Penalty

Be­ing leg­ible to other agents by com­mit­ting to us­ing weaker rea­son­ing systems

Me­ta­math­e­mat­ics and probability

Me­ta­math­e­mat­ics and Probability

Den­sity Zero Exploration