The Download: Why AI Breakthroughs and Fears May Be Overhyped

Summary

This edition of The Download argues that the latest wave of AI claims about hacking, mathematical breakthroughs, and self-improving superintelligence may be more hype than reality. Timnit Gebru and Emily M. Bender say that companies have strong commercial incentives to overstate capabilities, while the appearance of speed and urgency can distract policymakers and the public. The newsletter also reports that 22 nations have called for a new global body to oversee AI, including pre-deployment testing and common safety standards, although the United States and China did not join the declaration. OpenAI has separately proposed global AI safety standards and wants the United States to lead the effort. In the United States, Texas has paused new data-center permits pending a grid audit, while California has adopted laws covering data-center power and water use. AMD has become the twelfth $1 trillion company, with AI-chip sales helping its shares rise more than 180% this year. Meta’s AI agent Muse has reached the top of the US App Store, but it faces a major zero-day vulnerability, and Amazon has blocked it from shopping on its site.