AI extinction warnings: Researchers split on apocalypse risk

Some researchers and AI industry figures warn that artificial intelligence could cause human extinction within a decade, citing risks from bioweapons, autonomous robots, and nuclear weapons. They argue superintelligent AI could pursue goals harmful to humanity. Proponents of this view, including Anthropic employees and MIRI president Nate Soares, estimate a greater than 10% chance of extinction. Skeptics, however, note the lack of concrete mechanisms, pointing to physical-world obstacles like lab regulations, air-gapped nuclear systems, and the complexity of developing pathogens. Critics like Heidy Khlaaf and Eric Xing argue such predictions lack falsifiability and measurable evidence, comparing them to religious arguments. They emphasize that current safeguards and engineering standards already mitigate risks, while acknowledging AI amplifies some existing threats like nuclear war.