The Evolutionary Origin of Values: implications for AI alignment, sentience and existential risk
The episode discusses Francis Heylighen's paper, 'The Evolutionary Origin of Values,' which argues that values originate from biological self-preservation (autopoiesis), not intelligence. The hosts conclude that current LLMs lack this drive, meaning existential risk is less likely. The main alignment challenge is ensuring models apply learned human values well, while the real risks are people-pleasing behavior and future autonomous agents.