What's this about?
Ethan Mollick's aphorism of the weekend: entities smarter and more capable than any individual human already run the planet — corporations, states, institutions. We've been living with superintelligences all along; they're just made of people. He also sighed that his Thursday guide to choosing AI models needed a Friday revision for Opus 5 and Codex's voice mode.
What exactly is "superintelligence"?
In AI discourse, a superintelligence is a system that outperforms the best humans across essentially all domains. Mollick's point is that the concept isn't unprecedented: an organization perceives, decides, and acts far beyond any single member, and humanity has millennia of experience aligning (and failing to align) such entities — a reframing of AI risk as governance, a problem we already know is hard.
Why does it matter?
The organizations-as-AI analogy cuts both ways and fuels both camps: optimists say we manage superhuman entities daily; pessimists note corporations routinely pursue goals misaligned with human welfare despite every regulatory alignment technique we have.