我們要如何駕馭超級人工智慧，並同時避免類似「被機器人全面掌控」的災難發生？當我們還在為創造出全能機器人而努力時，人工智慧專家斯圖爾特 · 羅素（ Stuart Russell）已經開始投入不同的領域：機器人內在的不確定性。來聽聽他對能與人類和諧共處的 AI 的願景，這些 AI 通過常識、利他主義以及其他人類價值觀來解決問題。

人工智慧如何讓我們更美好

Superintelligent machines could make decisions beyond human capability, yet this rapid progress also brings risk. Giving AI fixed goals—like fetching coffee at all costs—can lead to catastrophic results, echoing the King Midas myth and the cautionary tale of HAL 9000. The solution is not to halt AI, but to design systems that act with humility, uncertainty, and a deep alignment to human values. Rather than blindly following orders, these machines must learn from human choices and remain open to correction.

Smarter Robots Need Uncertainty To Protect Humanity

"Slaughterbots"

code

communication

algorithm

humanity

machine learning

innovation

data

future

computers

robots

technology

science

decision-making

ethics

TED Stage Talk

斯圖爾特 · 羅素: 人工智慧如何讓我們更美好