(Overtime audio available to paid subscribers through the player at the top of the page and video version available below the paywall.)
1:29 Nora’s low score on the AI doomer test
7:58 The messy math of AI alignment
13:13 Teaching and (unteaching) concepts to LLMS
17:38 Do we really want human-like AI?
28:20 How the AI “mind” works
40:13 How to change an AI’s mind
52:51 Are LLMs reverse-engineering the human mind?
56:58 Heading to Overtime
Robert Wright (Nonzero, The Evolution of God, Why Buddhism Is True) and Nora Belrose (EleutherAI). Recorded January 16, 2024.
Twitter: https://twitter.com/NonzeroPods
Overtime titles:
0:00 Eleuther’s extra open open-source AIs
4:29 Is the human mind a general learning machine?
7:18 Why Nora supports open-source AI
11:42 How could we tell if AIs are conscious?
23:31 The new contender for reinforcement learning supremacy
29:57 Inside the minds of the AI doomers
40:52 Bob’s main sources of AI anxiety
