决策理论学家埃利泽·尤德考斯基（Eliezer Yudkowsky）带来了一条简单的信息：超级智能 AI 可能会杀光我们所有人。那问题就是：有没有可能做出强大的人工智能，而它会是顺从，甚至善意的？在这场尖锐的演讲中，尤德考斯基探索了为什么我们得立即采取行动，避免比人类更聪明的 AI 系统把我们送上灭绝之路。

超级智能 AI 会终结世界吗？

Powerful AI systems are being built faster than anyone can fully understand or control them. Current methods rely on tweaking vast, inscrutable networks until they work, with no real plan for aligning them to human values. Without a proven way to make superintelligent AI safe, even a single misstep could have irreversible consequences. The speaker warns that humanity is not taking this danger seriously and is running out of time to act.

Superintelligent AI Poses an Existential Threat We Ignore

technology