Some AI news for your Mean Green consideration

[Greenrex]

A partial quote from an article by Ted Nguyen in today's The Athletic Plus: Behind the AI magic that lets Amazon’s Prime Vision show the NFL like never before

"Find any activity that even remotely involves technology, and you’ll also find someone fretting about artificial intelligence. AI is going to make our lives so much better … or take over the world.    Those are worries for outside this newsletter. AI is, however, creeping onto our football broadcasts. And as Ted Nguyen writes today, the results are … eerie.  There’s a great opening story about Sam Schwartzstein, a former offensive lineman and one of the brains behind Amazon Prime’s Defensive Alert AI system. He thought it was buggy at first, before realizing it was highlighting blitzers even he hadn’t yet recognized. And no one really knows how the system identified that blitzer, either.   The AI, like any other machine learning tech, is ingesting data, in this case thousands of plays and copious tracking data, and learning how to predict future plays based on it. Simple concept, but the possibilities are wild, even in football. The viewer sees an in-depth look at what the quarterback is processing — correctly or incorrectly. It’s an intense watch, which I think is the point. Playing quarterback is hard.  My brain goes straight to: Who could this AI beat? Especially after more weeks of learning? Once AI learns how to call plays, I bet it could take an average Big Ten defense and hold Iowa to under 25 points." Paywalled  (I think) https://theathletic.com/4969578/2023/10/18/thursday-night-football-amazon-prime-vision/

 

[UNTLifer]

Who's Al?

[VideoEagle]

12 minutes ago, UNTLifer said:

Who's Al?

He did play quarterback in high school! 

[Udomann]

1 hour ago, Greenrex said:

My brain goes straight to: Who could this AI beat? Especially after more weeks of learning? Once AI learns how to call plays, I bet it could take an average Big Ten defense and hold Iowa to under 25 points."

Fun fact about AI: Surprisingly easy to confuse it in it's training phase. We've been trying to implement some predictive modeling with our reporting and stress testing it's capabilities.

Feed it a sufficiently complex algorithm where multiple results are possible, then tell it it is wrong (I.e. Chatgpt). Continue to do so from multiple sources and the AI will start feeding false data. So unless you have an enclosed AI system (which defeats the purpose of training from multiple sources), you can probably feed it a ton of false data/plays/reads and produce ineffective results. And let's be honest, at the revenue generation level of the NFL, people will be feeding that system every poor madden scenario possible.

Crap goes in and crap comes out.

Is AI useful? Sure. Is it coming for your job? yeah...not there yet. 

Edited October 18, 2023 by Udomann