Mobileye· @Mobileye · X·· 2 小时前
事实摘要
Mobileye 的 Amnon Shashua 与 Shai Shalev-Shwartz 发表了一项关于 AI 对齐的新研究，关注训练目标上的更好表现能否转化为实际期望的行为。在自动驾驶语境下，论文提出一个数学框架，用于研究 AI 能否从训练要求泛化到未见过的要求，包括罕见场景中的安全与舒适驾驶。
正文 · 原文
Amnon Shashua and Shai Shalev-Shwartz have published new research on a fundamental challenge in AI—alignment, ensuring that better performance on training objectives translates into the behavior we actually intend.
In self-driving, this means asking whether better imitation of human drivers translates into safe, comfortable driving—including in rare situations.
Their paper introduces a mathematical framework for studying whether AI can generalize from training requirements to unseen requirements.
Shai explains below.
@AmnonShashua @shai_s_shwartz
In Disney’s Fantasia (1940), Mickey Mouse enchants a broom to carry water. The broom does exactly what it’s told. But things quickly spiral out of control. This 85-year-old story captures one of the deepest unsolved problems in AI: alignment. 🧵在 X 查看被引用的帖子
来源：Mobileye · x.com
所属事件：Mobileye发布AI对齐数学框架研究