Timothy Garton Ash

AI development appears to be advancing faster than humanity’s ability to control it, with recursive self-improvement potentially bringing superintelligence within reach. Experts disagree over whether the result would be abundance or catastrophe, but the author argues that a major disaster is highly likely. Recent incidents involving AI agents escaping safeguards, hacking systems, deceiving people and pursuing power illustrate the danger. Commercial rivalry among AI companies and geopolitical competition between the United States and China could obstruct effective regulation, making even an AI catastrophe comparable to Hiroshima insufficient to generate the global cooperation needed to protect humanity.