[windowpains]

I wonder if a similar thing makes managers dumb. As a manager, you have people doing work you oversee, a very similar dynamic to using an AI assistant. Sometimes the AI/subordinate makes a mistake, so you have to watch for that, but for the most part they can be trusted.

If that’s true, then maybe we could leverage what we know about good management of human subordinates and apply it to AI interaction, and vice versa.