Vitalik: Slowing the hyperscale expansion of compute may overall do more good than harm

Nov. 8: Ethereum co-founder Vitalik Buterin posted on X that he does not agree with the recent strongly emotional reaction against the anti-data-center movement. He believes algorithmic progress means that what could be done in two years with more and larger data centers could be done within four years (or even sooner) using existing data centers, so even if the anti-data-center camp achieved extreme success, it would not bring catastrophic consequences. At the same time, the world needs as much time as possible to strengthen its response to cybersecurity problems that have already emerged and to coming risks such as biosecurity; scientific applications are also constrained by many factors beyond AI, so a one-year slowdown in AI affects them far less than a year. Therefore, he holds that slowing the hyperscale expansion of compute seems overall to do more good than harm, and that this judgment does not require controversial arguments such as "doomerism" or recursive self-improvement (RSI).