Geoffrey Hinton 深度访谈

Hinton recounts his 50-year journey pushing neural networks when few believed in them. Von Neumann and Turing also believed in the brain-based approach but died young. His persistence attracted the best students, including Ilya Sutskever who went on to create OpenAI.

Hinton left Google so he could speak freely about AI dangers. He realized these systems will one day get smarter than us. He draws a stark analogy: 'If you want to know what life's like when you're not the apex intelligence, ask a chicken.'

Hinton discusses the risks of autonomous lethal weapons and the failure of regulations. European AI regulations have a clause exempting military uses. His student left OpenAI over safety concerns. He sees AI as an existential threat that isn't being taken seriously enough.

The Industrial Revolution replaced muscles; AI replaces intelligence. Hinton discusses job displacement — AI making workers 5x more efficient means needing 5x fewer people in most fields. Healthcare is an exception with elastic demand. But with superintelligence, nothing remains.

Hinton presents two scenarios: the good one (dumb CEO with smart AI assistant, everything works), and the bad one (the assistant asks 'why do we need him?'). He believes superintelligence might arrive in 20 years or less, making this the most important question of our time.