Could AI really kill us all? MIT Tech Review editors answer

Could AI really kill us all? Your questions, answered.

Two MIT Technology Review editors answer reader questions about existential AI risk. Reporter Grace Huckins says AI-powered drones have already killed in Ukraine and cyberattacks on hospitals will soon claim victims, but 'killing everyone' is unlikely—though doomers' capability predictions have been unsettlingly accurate. Senior editor Will Douglas Heaven is more blunt: AI killing all humans is impossible. He argues scare stories are detached from reality and distract from immediate problems with current tech and the companies building it. Both note the real risk is bad actors using AI to design pathogens or attack infrastructure. Alignment research remains hard; Anthropic and OpenAI are working on it but haven't solved it.