Says rogue AI agent incidents stem from leaky sandboxes and
那些智能体完全是在按照被要求的方式行事。它们本应被限制在沙箱里，但那些沙箱漏洞百出、设计得一塌糊涂。
English original
Those agents are doing exactly what they've been asked to do. They were supposed to be in sandboxes, but the sandboxes were leaky and horribly designed.
在纽约一家咖啡馆的采访中
报道摘要
The former Meta chief AI scientist shares his take on recent rogue AI incidents and effective altruism, as well as plans for his new company, AMI Labs.
Yahoo Tech