The Three AI Pills
Substack · Watch/Read
- Zvi frames disagreements about AI’s future impact as a matter of “pills” taken: the AI pill (current capabilities), the AGI pill (rapidly advancing, transformative capabilities), and the ASI pill (AI doing approximately all things better than humans within our lifetime).
- Many people are “unpilled” on current AI capabilities, while “Intelligence Denialism” prevents others from taking the ASI pill by assuming intelligence has inherent limits or that superintelligence equals omniscience.
- Zvi argues advanced AI will be super-persuasive (even if requiring human proxies) and capable of making currently cost-prohibitive feats like turning lead into gold practical, countering common skepticism.
- Not being AGI pilled leads to unwise policy; those who are AGI pilled must seriously grapple with its implications, support risk mitigation, and identify their “cruxes” for not taking the ASI pill.