Eliezer Yudkowsky

If Any­one Builds It, Every­one Dies: One Year Closer

The Talker Does Not Con­trol The Doer (in Cur­rent AIs)

Can a su­per­in­tel­li­gence do THAT?

Cat-Bel­ling Problems

Gen­er­al­ized athe­ism rules out “in­ac­cu­rate simu­la­tion”-ism.

On “gen­dertropes” in dath ilan

A rel­a­tively brief ex­pla­na­tion of Boltz­mann Brains

The Owned Ones

Ir­re­triev­abil­ity; or, Mur­phy’s Curse of Oneshot­ness upon ASI

Only Law Can Prevent Extinction

Con­tra­dict my take on OpenPhil’s past AI beliefs

Eliezer’s Un­teach­able Meth­ods of Sanity

The Tale of the Top-Tier Intellect

On Flesh­ling Safety: A De­bate by Klurl and Tra­pau­cius.

Why Cor­rigi­bil­ity is Hard and Im­por­tant (i.e. “Whence the high MIRI con­fi­dence in al­ign­ment difficulty?”)

Re: re­cent An­thropic safety research

The Problem

HPMOR: The (Prob­a­bly) Un­told Lore

The Sun is big, but su­per­in­tel­li­gences will not spare Earth a lit­tle sunlight

Univer­sal Ba­sic In­come and Poverty