Import AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy
AI swarm scaling increases intelligence explosion risk due to parallelization despite diminishing returns
“I'd hoped that the value of λ for AI agents would be lower, making an intelligence explosion less likely, but that appears to not be the case”— Jack Clark
Toby Ord's analysis of AI swarm scaling finds that multi-agent swarms trade token efficiency for wall-clock speed, but exhibit diminishing returns analogous to human team coordination costs. More concerningly, swarm scaling appears potent enough to increase rather than reduce the probability of a recursive self-improvement-driven intelligence explosion. A secondary signal: 61% of Americans find voluntary AI self-governance commitments insufficient, directly challenging the Trump administration's industry-led approach.