MichaelDickens

Notes on the pos­si­bil­ity of moral progress

Paus­ing AI at hu­man level seems harder than paus­ing ASAP

Train­ing AI to be bet­ter at cor­rect­ness than persuasion

AI will make biolog­i­cal ex­tinc­tion risks worse be­fore it makes them better

A fron­tier AI com­pany should shut down

How valuable are weak AI safety reg­u­la­tions?

We Need Breadth-First AI Safety Plans

Sen­tient Welfare Across Three Futures

Can AI make ad­vance­ments in moral philos­o­phy by writ­ing proofs?

Paus­ing AI Is the Best An­swer to Post-Align­ment Problems

By Strong De­fault, ASI Will End Liberal Democracy

Which types of AI al­ign­ment re­search are most likely to be good for all sen­tient be­ings?

If AI al­ign­ment is only as hard as build­ing the steam en­g­ine, then we likely still die

Where I Am Donat­ing in 2025

Align­ment Boot­strap­ping Is Dangerous

We won’t solve post-al­ign­ment prob­lems by do­ing research

Know­ing Whether AI Align­ment Is a One-Shot Prob­lem Is a One-Shot Problem

Epistemic Spot Check: Ex­pected Value of Donat­ing to Alex Bores’s Con­gres­sional Campaign

Things I’ve Be­come More Con­fi­dent About

Out­live: A Crit­i­cal Review