March Madness Puts Machine Learning and AI to the Test

Indication of progress or more hype?

Whether or not you follow NCAA basketball during the season, you’re drawn to the adrenaline rush of March Madness because it’s all about gambling. The odds of filling out a perfect bracket (predicting 63 games correctly) ranges from 1 in 128 billion to 1 in 9.2 quintillion.

This isn’t the first year that companies are using the tournament to drum up machine learning and AI hype but here are a couple 2018 examples:

• Google and the NCAA launched an AI competition with $100,000 on the line for the teams with the most accurate bracket prediction models, fueled by data about past team performance. They’ll likely do another media push once the winners are announced so stay tuned!
• Unanimous A.I., an SF-based tech company, leveraged collective “intelligence” or “swarm A.I.” to fill out a bracket. You can see where the hive has messed up so far here.

Is data about the past a good indicator of what will happen in the future? Is crowd-sourcing better than individual blind guesses? The Virginia Cavaliers, after a 31–2 season, lost in a blowout, to the University of Maryland, Baltimore College Retrievers. Until then, a 16 seed had never upset a 1 seed in the tournament, in 135 games. In a historical upset, game 136 went differently.

I fully support the educational benefit of these challenges but it’s hard to deny the larger role that stunts like this play in increasing the ambiguity around what machine learning and AI actually are, how this technology is being used now and will be used in the future, and what it means society. My takeaway is we should leave March Madness to luck and publicize AI applications that can solve real-world problems and conversely, discuss how we can prevent bad actors from using AI maliciously.

I’ll be sure to share balanced and informative articles when I find them. In the meantime, enjoy the tournament!