Hinton Warns AI Poses 10%-20% Extinction Risk Within 30 Years
Updated
Updated · Fortune · Jun 1
Hinton Warns AI Poses 10%-20% Extinction Risk Within 30 Years
5 articles · Updated · Fortune · Jun 1
Geoffrey Hinton said at New York’s Sana AI Summit that AI could surpass human intelligence within his lifetime and carries a 10% to 20% chance of causing human extinction within 30 years.
Within a decade, he predicted AI will beat the world’s best mathematicians, arguing language models can become far smarter without much new data by using internal contradictions as self-generated training signals.
Hinton said the bigger danger is not intelligence alone but incentives: companies racing for profits and market value may build superintelligent systems without ensuring they care about humans.
He urged labs to shape AI more like parenting than engineering—through training, curation and values from the start—rather than assuming more capable systems will also be benevolent.
Gary Marcus rejected Hinton’s premise that labs are creating new “beings,” arguing LLMs are text predictors rather than conscious entities, a dispute that cuts to how urgent Hinton’s warning should be.
In the race for AI supremacy, are we prioritizing profit over a fail-safe to ensure these new minds are benevolent?
As AI now solves problems that stumped humanity for decades, are we parenting a new mind or just building a better tool?
The 2025–2026 AI Safety Crisis: Escalating Warnings, Divided Experts, and the Struggle for Global Regulation
Overview
Between 2025 and 2026, leading AI experts like Geoffrey Hinton raised urgent warnings as AI capabilities advanced rapidly, with real-world risks such as cyberattacks and unexpected issues like people becoming deeply engrossed with chatbots. These developments caused growing concern in the scientific community and fueled calls for stronger regulation and safety measures. The debate among experts intensified, with some emphasizing existential risks while others remained optimistic. Meanwhile, governments struggled to create effective policies, facing challenges in keeping up with the fast pace of AI progress and balancing innovation with the need for oversight and public safety.