179 – The Plan (to align AI), with John Wentworth

Worried about AGI running amok in the near future? John Wentworth thinks we can align AI. In 10-15 years. With greater than 50/50 probability. And he has a plan! We discuss The Plan, its merits, and how it has progressed over the past year. Primary Sources: The Plan The Plan – 2022 Update Also discussed: The Basic Foundations for Agent Models sequence The Telephone Theorem The “Minimal Latents” Approach to Natural Abstractions Help With The Plan, Get The Skills, Save The World: Read The Embedded Agency Sequence Join SERI MATS! (see also SERI MATS tag on LessWrong) Apply for funding from The Long-Term Future Fund 56:05 – Guild of the Rose Update 57:36 – Feedback 58:20 – LW posts 1:19:09 – Thank the Patron We now partner with The Guild of the Rose, check them out. Hey look, we have a discord! What could possibly go wrong? Our Patreon page–your support is most rational, and totally effective. (also merch) Rationality: From AI to Zombies, The Podcast LessWrong Sequence Posts Discussed in this Episode: Expecting Beauty Is Reality Ugly? Next Episode’s Sequence Posts: Beautiful Probability Trust in Math