AI Talk: Mercedes AVTR, pass completion probability

October 2nd, 2020 / By V. “Juggy” Jagannathan, PhD

Normally, I don’t comment on marketing material, but these two stories are a bit out of the ordinary.

Mercedes AVTR

In Futureloop’s email digest I saw a reference to a new concept car released by one of the luxury car manufacturers. I took a quick peek at the YouTube video about the car and was simply blown away!

Inspired by the science fiction movie Avatar, they have managed to create science reality! The car has been built with intelligence at its core: Electric (of course), made out of recyclable plastic and other biofriendly material and unbelievably sleek with “DeLorean” doors. There is no steering wheel! Instead there is a control unit between the front seats that either person can operate by simply palming a the device. The car can move laterally as well with crab-like wheels.  Plenty of AI to go around. Your palm becomes a display device and you interact with the car using gestures and touch. Open an app with voice and close it with a gesture. Colorful lights are engineered to display emotion. A battery of fin-like devices in the back of the car, open, shut and sway. It is supposed to convey the driver intent: Is the driver accelerating, slowing down, or turning to one side or the other? There are a number of interviews in the video—the head User Experience designer, who goes into the motivation for creating the car and the designer who came up with the body styling in an internal competition. Indeed, a futuristic car. Wonder how far into the future one needs to wait to actually see these on the road?

Pass Completion Probability

All the major league sports are now using AI in some form or another. In fact, machine learning has a long history in baseball. Moneyball, the best-selling book by Michael Lewis, chronicles the use of data-driven insights to build a competitive Oakland Athletics team in 2002. Fast forward to 2020, what is the NFL up to in this space? This case study on machine learning published by Amazon is an interesting read. One of the fundamentals of being able to use machine learning is availability of data. So how are they collecting the data and what data are they collecting? Turns out they are awash with all kinds of data. There are RFID chips in the footballs and the shoulder pads of players! The signal from these chips is collected from RF receivers around the stadium. They get video feeds as well. Lots of processing goes into manipulating all this data—and that’s where the Amazon cloud and machine learning (ML) solutions factor in.

One of the metrics that this new platform is able to predict is “pass completion probability.” It stands to reason shorter passes have higher probability of completion than longer ones. But how different are they actually? Currently, quarterbacks are not given credit based on the difficulty of a pass that is completed – simply aggregate yardage. The ML team literally use 10-different factors to compute this probability. The intent is to provide all this data in real-time to commentators to change the narrative and user experience of watching the game. This is just one of the stats and, it is still fairly early for NFL use of intelligent stats. Scores of stats are being computed and the commentators are going to have a field day when all of this becomes common place. And, nerds like me (a big Steelers fan), would like to guess at Big Ben’s pass completion probability from the couch.

Acknowledgements

The NFL story above was sent to me by my friend, Chandy.

I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something!” Leave me a comment below or ask a question on my blogger profile page.

V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.