Robbins 支持对模型发布和护栏设立合理规则,但认为 AI 的负面影响与以往技术一样可以通过技术手段解决。
我们当然需要对模型如何发布、护栏设置有一套逻辑,我们应当了解发生了什么。但我们引入的每一项技术都有正反两面,你只需要解决负面问题。原文 (en)
We certainly need to have logic around how models get released, logic around guardrails. We should know what's going on. But every technology we've ever introduced has had positives and negatives. I mean, you just have to solve for the negatives.