IBM Serves Up Its AI Progress at Wimbledon 2019

We go backstage at the Grand Slam tournament

I recently had the privilege of a behind-the-scenes tour of IBM’s latest artificial intelligence (AI) technology being deployed at the 2019 Wimbledon tennis championships. When it comes to AI, many of the public cloud providers seem to attract all the attention. And in an enterprise market that’s largely experimenting with the technology, I’ve been impressed by the range of IBM’s AI customers in 2019, and most of all, by how the company quietly goes about enabling them to fully operationalize the technology.

Big Blue is Wimbledon’s main technology partner and since 1995 has been providing a host of IT infrastructure, hybrid cloud, cybersecurity and real-time digital solutions for fans, players and broadcasters. As an example of the depth of this partnership, this year IBM had over 180 dedicated staff on site to support the IT infrastructure needs of the three-week tournament.

Wimbledon has also been using IBM Watson since 2015 and has seen AI flourish in many areas of the event, spanning chat bots, media and content, social media analytics and cybersecurity. This year, however, Wimbledon took a big leap forward with its AI deployment.

Let’s take a look at some of the initiatives in play and how they reflect key aspects of the market.

Wimbledon’s Big Leap in AI

In 2017, Wimbledon’s efforts in AI went up a level with the unveiling of the Wimbledon Cognitive Highlights Solution, a cloud-based system designed by IBM to accelerate the creation of highlights packages for fans and for broadcasters. Back then, IBM’s cloud enabled the mapping and collecting of video footage, while a set of the company’s machine learning and deep learning algorithms running over the content helped Wimbledon’s editors decide which clips would make the best five-minute highlights video of the match.

Fast-forward two years and in 2019 the solution has been fine-tuned, taking real-time video feeds from 10 different courts, each of which hosts up to four matches every day. Additionally, the solution now combines visual analysis through IBM Watson Visual Recognition, a set of APIs that allow for accurate tagging, classification and training of visual content using IBM’s machine learning, and for audio analysis through IBM Watson Acoustics. This tool assesses scoring data, player gestures and crowd noise to determine an overall excitement score for each scene within a match.

Once a match is completed, the system automatically generates highlights packages based on the excitement scores for the most gripping scenes, cropping the clip tightly through acoustic analysis of when a ball has been hit. The packages are then shared with Wimbledon’s digital editors, who select which ones to use to share on the tournament’s online channels. When the system was first rolled out a year ago, it resulted in a more than 250% increase in the number of highlights packages, and in higher-quality highlights throughout its channels. This video content attracted over 14 million net new views in 2018.

In Watson Goes Anywhere at IBM Think 2019, I assessed IBM’s strategy in AI and argued that despite some challenges, the company has some differentiated and competitive enterprise services for implementing AI into operational processes. The level of operational improvement Wimbledon is displaying in its deployment of Watson further backs up this view.

Tackling Bias with IBM Watson OpenScale

The most interesting aspect of Wimbledon’s AI deployment this year, however, was its use of IBM Watson OpenScale to help it tackle bias in the process of creating highlights. OpenScale is IBM’s AI management and governance platform launched in 2018 that encompasses bias, performance monitoring, explanation and transparency tools for machine learning — for my views on this solution, see Trusting Watson.

Addressing bias in data has become a top customer concern about machine learning over the past 18 months and is the main motivation behind IBM’s launch of OpenScale, which remains a differentiated capability in the AI space. The platform addresses fears over “black box” AI by giving customers ways to improve the visibility and control of systems, consistently check for bias and improve the interpretability of outcomes. It does this by monitoring the performance, health and behaviour of machine learning models at runtime.

Although Wimbledon’s AI-driven processes dramatically accelerated the time to assemble highlights packages from previous years, event organizers still needed specialists to manually identify potential bias in excitement scores, for example when matches taking place in larger courts invited bigger crowds, or when a player had a particularly vocal set of fans. To overcome this, for this year’s tournament IBM introduced new bias monitoring using OpenScale to automatically identify possible bias in each scene, taking into account the court of play and each player’s ranking.

In today’s real-time media environment, the level of automation this year meant that video packages were able to be completed, checked for bias, within two minutes of a match finishing — a full 15 minutes faster than in previous years.

A Great Showcase for the Potential of AI

Impressive as these improvements are, IBM’s efforts at Wimbledon above all are a great public showcase of the potential that AI holds, especially when deployed at such an operational level. Too often I see AI projects still in experimental phases in organizations. Or, when the technology is deployed in a production environment, the benefits often go unnoticed by the wider community because AI is invisible and works behind the scenes, like it does at Wimbledon.

It was great to see IBM and Wimbledon showcase the work they’re doing. My tour left me feeling very positive about where the two partners will take AI in the future and I’m already looking forward to seeing what advances next year’s tournament will bring.

I should also point out that it wasn’t all about AI — the tennis I saw wasn’t so bad either.

Wheelchair tennis match at Wimbledon 2019