Both Sides Are Overconfident On AI | Toby Ord

Swarm scaling offers a mechanism to accelerate AI capabilities by parallelizing tasks across multiple agents, though it introduces a trade-off between compute costs and time efficiency. This relationship is defined by a power law exponent, lambda, which determines how effectively additional agents translate into performance gains. While some tasks, such as knowledge base creation, exhibit near-perfect parallelization, others suffer from "stepping on toes" where agents interfere with one another. Beyond technical scaling, the discussion addresses the uncertainty surrounding AI timelines and the potential for recursive self-improvement. Toby Ord, a senior researcher at Oxford’s AI Governance Initiative, notes that while existential risks remain a serious concern, proactive governance and international cooperation provide a path to mitigate these dangers. Effective altruism, rather than being a fatalistic movement, focuses on using reason and evidence to identify high-leverage interventions to secure humanity's future.