Golf.com is reporting that IBM Watson will be used at the Masters golf tournament this weekend to identify activity that merits the spotlight of television and online video feeds. Why would you need to do this? Aren’t the great shots obvious? This AI application goes well beyond optimizing a single television broadcast sorting through which of 20,000 golf shots to highlight over four days.
The Future of Personalized Television Broadcasts Courtesy of AI
The issue isn’t whether it is known what shots are good. It is instead about how you collect that information in one place, tag it and prioritize it for the main audience and eventually for individual audiences. This is an AI application that on the surface is about optimizing a television broadcast to highlight the most exciting and expert golf play, but points to a future of personalized viewing that no human curation team could hope to do in real-time. An interview by Golf.com with IBM’s John Kent sheds more light on the future potential for AI in content curation:
‘When you think about the golf that’s going on with the number of golfers and number of cameras out there, it affords the Augusta National editorial team the ability to scale up,’ John Kent, the head of IBM’s tech approach at the Masters, said. ‘I could do a highlight real for Shane Lowry, every shot of his, very easily. I could put a package together of your favorite players. I could create an experience for an individual coming to the site that is much more personal. I could have a recap video when you come on that is just ‘Hey, here’s what happened the last 15 minutes.’ The possibilities are endless.’
Using AI to Track 20k Golf Shots
This is a task that would be very difficult to address with humans. There are generally about 90 golfers in the first two rounds of The Masters played on Thursday and Friday. The field is then cut to the top 50 scores for the final Saturday and Sunday rounds. It is usually a little more than 50 golfers on the weekend because they include everyone tied with the 50th place player. If all of the golfers averaged the 72 par for their rounds, that is more than 20,000 golf shots over the weekend or more than 10 every minute during the peak times. Of course, not everyone shoots par so the numbers are even larger.
You would need a lot of humans to keep up with this volume and then the question comes to judgment on which shots deserve priority for television air time. The scoring leaders get priority, but so do popular players or particularly good shots. In the future, this might deliver personalized compilations of a fan’s favorite golfer.
IBM Watson Will Evaluate Sight, Sound and Words
The question then is what data should the AI be evaluating to determine what activity deserves priority. IBM has come up with an “Overall Excitement Score” which adds up three factors:
- Crowd Cheering / Commentator Excitement
- Action Recognition
The crowd cheering and commentator excitement requires IBM Watson to understand a baseline of activity such as before a shot and after a shot to understand what normal and heightened levels of excitement are. Site and sound will be key inputs. Similarly, it will need to gauge commentator speech patterns, speaking volume and other emotional characteristics to assess commentator excitement. Action recognition can be derived from visually reviewing activity of the golfers, the equipment and crowd. Commentary requires an understanding of natural language. Golf.com’s Sean Zak put it this way:
Things like player reactions, where Watson knows the difference between a fist pump and a simple tip of the cap, crowd cheering where there’s a clear difference between the common golf clap and the patented Augusta roar, and even commentary from broadcasters. Yes, if Peter Kostis declared an approach “phenomenal,” Watson would know that it wasn’t just good, it was indeed phenomenal.
This is complicated stuff. Much of the same technology used in self-driving cars and natural language understanding is required to sort through copious volumes of unstructured data. This is an example of computers that are developing a way to understand of activity in the natural world.
A Tool That Spans Healthcare and Entertainment
The Masters application of IBM Watson helps illuminate the breadth of use cases where powerful AI solutions can add value. Optimizing a televised sports broadcast may not deliver as much societal value as helping doctors diagnose cancer, stopping cyber attacks or improving energy consumption efficiency, but versatile technologies are important because they can go beyond a single application. It will be interesting to see if Masters coverage will be better this year. I’ll be watching with extra interest.