The Download: controlling AI agents and “Super Intelligence” loyalty tests
| Source: MIT Technology Review AI
Tags: MIT Technology Review, Reverse game theory, Nvidia, Donald Trump, US midterms, Export controls, AI agents
MIT Technology Review's daily briefing covers reverse game theory as a way to keep AI agents within rules, the Trump order rebranding AI as "Super Intelligence" turning into a loyalty test, and Nvidia telling data centre clients to cap China business as US midterms test AI's political weight.
Details
Reverse game theory flips the usual question: instead of predicting how rational actors behave under given rules, it designs rules that make the desired behavior the rational choice. Researchers hope this can keep powerful AI agents aligned with instructions, though it remains unproven in practice. A month after President Trump's executive order requiring agencies to rebrand AI as "Super Intelligence," the term is spreading across government sites, while many major tech companies still say "AI." The concern is that the label becomes a signal of political allegiance. The briefing also flags that Nvidia has told data centre providers that most of their sales must be non-Chinese, following pressure from Washington, and that today's US midterms will show whether AI regulation sways voters.