AI Godfather Geoffrey Hinton warns that We're Creating 'Alien Beings that "Could Take Over"

So will AI wipe us out? According to Geoffrey Hinton, the 2024 Nobel laureate in physics, there's about a 10-20% chance of AI being humanity's final invention. Which, as the so-called Godfather of AI acknowledges, is his way of saying he has no more idea than you or I about its species-killing qualities. That said, Hinton is deeply concerned about some of the consequences of an AI revolution that he pioneered at Google. From cyber attacks that could topple major banks to AI-designed viruses, from mass unemployment to lethal autonomous weapons, Hinton warns we're facing unprecedented risks from technology that's evolving faster than our ability to control it. So does he regret his role in the invention of generative AI? Not exactly. Hinton believes the AI revolution was inevitable—if he hadn't contributed, it would have been delayed by perhaps a week. Instead of dwelling on regret, he's focused on finding solutions for humanity to coexist with superintelligent beings. His radical proposal? Creating "AI mothers" with strong maternal instincts toward humans—the only model we have for a more powerful being designed to care for a weaker one.

1. Nobody Really Knows the Risk Level Hinton's 10-20% extinction probability is essentially an admission of complete uncertainty. As he puts it, "the number means nobody's got a clue what's going to happen" - but it's definitely more than 1% and less than 99%.

2. Short-Term vs. Long-Term Threats Are Fundamentally Different Near-term risks involve bad actors misusing AI (cyber attacks, bioweapons, surveillance), while the existential threat comes from AI simply outgrowing its need for humans - something we've never faced before.

3. We're Creating "Alien Beings" Right Now Unlike previous technologies, AI represents actual intelligent entities that can understand, plan, and potentially manipulate us. Hinton argues we should be as concerned as if we spotted an alien invasion fleet through a telescope.

4. The "AI Mothers" Solution Hinton's radical proposal: instead of trying to keep AI submissive (which won't work when it's smarter than us), we should engineer strong maternal instincts into AI systems - the only model we have of powerful beings caring for weaker ones.

5. Superintelligence Is Coming Within 5-20 Years Most leading experts believe human-level AI is inevitable, followed quickly by superintelligence. Hinton's timeline reflects the consensus among researchers, despite the wide range.

Keen On America is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit keenon.substack.com/subscribe