Geoffrey Hinton Warns of 10% Chance AI Kills Humans

Nobel laureate Geoffrey Hinton warned in a BBC Newsnight interview that there is a 10 percent chance artificial intelligence could kill all humans within the next decade. Hinton, who resigned from Google in 2023 to speak freely on these risks, noted that his estimated timeline for such a catastrophe has shrunk from 30 to 50 years down to 10 years or less.

Hinton argued that AI does not need physical presence to cause devastation, citing the potential for cyberattacks, the design of biological viruses, or the creation of chaos through communication. He urged engineers to design systems that do not want to harm humanity and called for global regulation to prevent AI from taking over.

These warnings coincide with the resignation of Jacob Coxon from Anthropic. Coxon accused major AI corporations, including OpenAI, of prioritizing development speed over safety and gambling with human lives in a race toward self-improving superintelligence. Evan Hubinger, alignment science lead at Anthropic, supported these concerns, stating he personally believes there is a greater than 10% chance AI could kill all humans within the next decade due to recursive self-improvement.

Other experts have provided varying estimates for this probability, known as p(doom). While Roman Yampolskiy estimates the risk at 99.99%, others like Elon Musk and Anthropic CEO Dario Amodei provide estimates ranging from 10% to 25%. Conversely, AMI Labs founder Yann LeCun argues that such a catastrophe is less likely than a nuclear holocaust.