Quintin Pope

Count­ing ar­gu­ments provide no ev­i­dence for AI doom

QAPR 5: grokking is maybe not *that* big a deal?

Re­search agenda: Su­per­vis­ing AIs im­prov­ing AIs

Evolu­tion pro­vides no ev­i­dence for the sharp left turn

Quintin Pope’s Shortform

My Ob­jec­tions to “We’re All Gonna Die with Eliezer Yud­kowsky”

A Short Dialogue on the Mean­ing of Re­ward Functions

QAPR 4: In­duc­tive biases

QAPR 3: in­ter­pretabil­ity-guided train­ing of neu­ral nets

Quintin’s al­ign­ment pa­pers roundup—week 2

Quintin’s al­ign­ment pa­pers roundup—week 1

The shard the­ory of hu­man values

Evolu­tion is a bad anal­ogy for AGI: in­ner alignment

Hu­mans provide an un­tapped wealth of ev­i­dence about alignment

[Question] What’s the “This AI is of moral con­cern.” fire alarm?

[Question] Any prior work on mu­ti­a­gent dy­nam­ics for con­tin­u­ous dis­tri­bu­tions over agents?

Idea: build al­ign­ment dataset for very ca­pa­ble models

Hy­poth­e­sis: gra­di­ent de­scent prefers gen­eral circuits

The Case for Rad­i­cal Op­ti­mism about Interpretability

[Linkpost] A Gen­eral Lan­guage As­sis­tant as a Lab­o­ra­tory for Alignment