UFC Jacksonville was fun, huh? Josh Emmett got beaten to a bloody pulp, Justin Tafa almost lost an eye, and something else happened… What was it… Oh yeah! PickGPT went 11/12 and came second in the world on the Tapology rankings! It’s alive! And whether it is in BETA or not, it stomped virtually all humans in predicting the outcomes of fights! Why? Because computers don’t have emotions or doubts, or even sexual attractions to cloud their judgments (looking at you Jacob). My computer, for instance, doesn’t get mad at me if I forget to send it flowers on the anniversary of the date I bought it. It doesn’t complain when I ask it to do something. It doesn’t piss or shit on my floor or tear up my carpet. I guess what I’m saying is that we should all strive to be a little more like computers.
We released PickGPT (@PickDoctorAI), a Neural Network AI that picks fights. It is only in Beta, but absolutely CRUSHED #UFCJacksonville! It went 11 for 12 in picks, was 2nd in the world on Tapology and is up almost 11 Units!
Follow it's journey at https://t.co/NoQzd6prAf pic.twitter.com/o8NwefphU2
Ok, now that I’ve gotten that little bit of bragging and asserting dominance over Jacob out of the way (anyone seeing a trend here), it’s time to let a little bit of the pressure out of these tires. I want you all to take a deep breath, and try to remember that 90%+ accuracy probably won’t be happening next week. To put it simply, PickGPT should not be this successful this soon, and while I try to figure out how and why that happened, let’s just bask in its glory for a few days and remember, HUMANS SUCK!!!! So far in 3 events, it has gone 7/10, 8/11, and 11/12. Most impressive is that each and every time it picked an underdog, that fighter won. Once you train a neural network like PickGPT, it can be very difficult to figure out what it is actually doing. The training is a result of some very advanced mathematics in a super high dimensional space. Because of this, once a neural network is trained, it often functions as a black box; information goes in, and information comes out, with no easily distinguishable explanation of how. Of course, that’s not going to stop the Pick Doctor from speculating!
So what explains the amazing performance on the UFC Jacksonville card? I have two theories. The first is that aside from two fighters, Chepe Mariscal and Joshua Van, all of the fighters had UFC data to feed in, and the only fight it got wrong was the Chepe Mariscal vs Trevor Peek fight. Second is that PickGPT was trained using biometric and striking data only, and UFC Jacksonville was a striking-heavy card. Even fights where I personally expected a lot of wrestling, had little or none of it. I expected Rebecki vs Radzhabov to be a wrestling match, and there was none. I expected Ricci vs Robertson to be contested on the ground, but was mostly contested on the feet. Since PickGPT does not yet know how to account for grappling stats, I think the striking heavy nature of the card helped quite a bit.
Of course, there is a possibility that it is just that accurate already, and 90% is a reasonable expectation going forward. I would caution you that if that were true, the enslavement of the human race to the machine overlords may be closer than we think. While some of us welcome our new masters with open arms, I have a feeling that may not be a popular sentiment, and AI taking your job may not be the most pressing issue on the horizon (unless you are Jacob). Stay tuned to follow the journey, and thanks for being a Premium member. You aren’t getting this kind of product anywhere else!
