Competition" has been said to be pointless, because the AI is always improving, but here's the thing: I don't need to be able to write a test that will defeat ChatGPT version 8. I only need a test that can defeat the current version, since I write new exams every semester. When I wrote my final this week, I ran each of the questions through ChatGPT and stopped when ChatGPT was scoring a 60% or worse on the test. Nonetheless, out of the 5 that have taken the final so far (it's early in the testing window), 2 of them used ChatGPT for the answers, and did poorly as a result. The best part is I don't even have to worry about confronting them or reporting them to academic integrity, as making a test this way results in them getting an F or a D anyway.

That's for my online classes. For in-person, I just switched back to paper after catching 20% of my class using ChatGPT on the online competency exams, and then a handful still using it immediately again right after I just caught them. Those people got sent to the dean, because at the point in which you get caught, have a conversation with me about it, and then cheat the same way again the next day... clearly nothing I am saying is getting through to you, so maybe the Dean's words will have more of an impact. Paper is annoying to grade, so I'm looking to figure out a better way going forward.

I sent more people to academic integrity this semester than I have, total, in the eight years I've been teaching full-time. Like Frank, I don't consider myself a disciplinarian, my main concern is that they actually learn the material, and cheating sabotages those efforts and hurts honest students. As what Frank found, when confronted, 100% of my students confessed, after initially denying it (and swearing that they would never cheat!). I usually can work with my students and just have a conversation with them and get them to knock it off, but ChatGPT is just so easy to use they use it anyway.