one complicated thought: out of the box/on the free tier, chatgpt is a dangerous product for the majority of users, in that it is capable of doing real harm, mostly invisibly. it takes a _huge_ amount of time and effort to resolve into a useful tool and the system basically fights you the whole way.

Computers should be able to do anything you tell them to but that process should be at least a little difficult.

In winter, I spoke at a faculty event about AI. I was alarmed by the range of faculty awareness. Some believe our students are so "good" they'd never cheat. Yes, cheating's wrong. But systems that explicitly or implicitly incentivize cheating--AI does both--will produce cheating. 11/