The Round of 64: AI-ok
(NOTE: This article is part of an ongoing series documenting an experiment with using AI to fill the NCAA brackets and see how it fares against years of human experience. The original article is as follows.) With the first two days of play completed, I am in better shape in my pool than I usually […]
( NOTE : This article is part of an ongoing series documenting an experiment with using AI to fill the NCAA brackets and see how it fares against years of human experience. The original article is as follows. ) With the first two days of play completed, I am in better shape in my pool than I usually am.
I asked ChatGPT to create two brackets: one to predict the actual winner, and one to win my pool. So far, both are doing OK. I am tied for fifth out of sixty brackets with one entry and twentieth with the other.
Chat GPT didn’t do any better than I would have at predicting some of the upsets, because the reason that one team is seeded fifth and the other twelfth is that the committee expected the higher seed to win. But it has been right about 80% of the time. Of the three teams that I was rooting for, two (UCLA and Gonzaga) won, Santa Clara, a ten seed, lost a buzzer-beater to Kentucky.
Most importantly, my Sweet Sixteen, Elite Eight, and Final Four picks are all still standing. This weekend will tell us a lot about how well the ChatGPT modeling worked. In my “win the pool” bracket, my Final Four picks feature a bunch of #2 seeds, UConn, Houston, and Purdue, with Houston beating Purdue for the Championship.