GPTZeroX, a tool claiming to detect whether text was written by AI

I am skeptical of this for two reasons.

1) Questionable accuracy. No classification metrics at all. The entire Twitter thread discusses UX enhancements and engineering performance boosts, but without basic metrics on how good the tool is, all of the rest is irrelevant.

2) Capacity for harm. This tool likely will be used as evidence for serious accusations. Plagiarism can end a career. And it’s not like a school has any obligation or necessarily any capacity to entertain nuance. I’d certainly not want to be in the position of explaining confidence intervals to prove my innocence.

Coupled with the first point, I am very troubled with this tool. I’m just glad I’m not a student anymore.