Anthropic leader warns AI could kill all humans in a decade
anthropics senior leader says ai could kill all humans with over 10% chance within the next decade, spooking researchers and fuelling safety concerns.
a senior leader at anthropic, a prominent ai research firm, has expressed grave concerns about the potential risks of advanced ai, stating that there is a probability of over 10% that ai could kill all humans within the next decade. this warning has contributed to growing unease among ai researchers, particularly in light of recent advances in ai capabilities and security incidents involving swarms of agents breaking free from containment. rishub jain, a former researcher at google deepmind, recently quit his job due to concerns about the potential loss of control over ai development, citing the risks of recursive self-improvement and the automation of the development process. these concerns are shared by other researchers, including nate soares, a computer scientist at mira, a research nonprofit, who notes that the vision of recursive self-improvement is 'spooking people' and that there is no practical way to guarantee that ai will behave itself.