avturchin

Side­load­ing: cre­at­ing a model of a per­son via LLM with very large prompt

If I care about mea­sure, choices have ad­di­tional bur­den (+AI gen­er­ated LW-com­ments)

Quan­tum Im­mor­tal­ity: A Per­spec­tive if AI Doomers are Prob­a­bly Right

Bit­ter les­sons about lu­cid dreaming

Three main ar­gu­ments that AI will save hu­mans and one meta-argument

De­bates how to defeat ag­ing: Aubrey de Grey vs. Peter Fedichev.

Magic by forgetting

Strength­en­ing the Ar­gu­ment for In­trin­sic AI Safety: The S-Curves Per­spec­tive

The Sharp Right Turn: sud­den de­cep­tive al­ign­ment as a con­ver­gent goal

Another for­mal­iza­tion at­tempt: Cen­tral Ar­gu­ment That AGI Pre­sents a Global Catas­trophic Risk

Run­ning many AI var­i­ants to find cor­rect goal generalization

AI-kills-ev­ery­one sce­nar­ios re­quire robotic in­fras­truc­ture, but not nec­es­sar­ily nanotech

The AI Shut­down Prob­lem Solu­tion through Com­mit­ment to Archiv­ing and Pe­ri­odic Restoration

Long-term mem­ory for LLM via self-repli­cat­ing prompt

Log­i­cal Prob­a­bil­ity of Gold­bach’s Con­jec­ture: Prov­able Rule or Coin­ci­dence?

A Pin and a Bal­loon: An­thropic Frag­ility In­creases Chances of Ru­n­away Global Warm­ing

The table of differ­ent sam­pling as­sump­tions in anthropics

Another plau­si­ble sce­nario of AI risk: AI builds mil­i­tary in­fras­truc­ture while col­lab­o­rat­ing with hu­mans, defects later.

Un­typ­i­cal SIA

Rus­sian x-risks newslet­ter May 2022 + short his­tory of “method­ol­o­gists”