OpenAI CEO Sam Altman described three potential endgame scenarios for artificial intelligence to staff last week, according to reports. The scenarios address how to safely manage the race to build superintelligence.
One path is an international AI regulatory body modeled on the International Atomic Energy Agency (IAEA). Another is a collaborative merger of leading AI labs into a single entity modeled on CERN.
The third scenario is that one company achieves recursive self-improvement and uses its power to “melt everyone else’s GPUs,” effectively dominating the field. Altman reportedly prefers the regulatory model and called the dominance scenario “terrible.”