Wimbledon and Watson, two legends of their respective fields, have teamed up to produce AI-powered tennis content which on paper has all the minerals to cook up a masterpiece. Yet the collaboration feels like it has fallen well short of what the pair of households names should be capable of, which we feel reflects how the market has reached something of a dead end with regard to applying AI to revolutionizing TV.
For this year’s tournament, simply put, Watson AI is producing highlight reels. Let’s be honest, highlights are nigh impossible to make sexy no matter how many buzzwords you tack onto the reel. Is spinning up highlights really all Watson AI – one of the world’s most powerful supercomputers – is capable of? Perhaps the problem isn’t in the technology itself, but in the demand from viewers and the cost savings which Watson can enable from the production end.
Despite sounding dull, there are some clever components of Watson’s highlight creation capabilities, so bear with us. Inputs include data as subtle as player’s facial expressions and emotional reactions, as well as analyzing the mood of the crowd and even tuning into acoustic data relating to the sound of the ball being struck which apparently aids with stitching together highlight frames automatically, i.e. identifying key shots and longer rallies. A key new element of this process is churning all this data while removing bias for the bigger courts hosting bigger players with bigger audiences – where the atmosphere will naturally be noisier than a low-key match.
“Not all highlights during a tennis match are equal. For example, a highly passionate crowd favorite could generate more excitement than a more reserved yet equally skilled opponent. As a learning system, Watson has been taught to better recognize acoustics and understand inadvertent bias increasing the quality of the output,” explained IBM.
Surely then the project is really more about automation as opposed to AI – two terms which are commonly confused. But Watson assures us that the project relied heavily on its one-year old product AI OpenScale, aimed at reducing bias in AI applications. Combined with Watson Acoustics, IBM is building highlight reels in a way we have seldom seen.
“Using Watson Open Scale, the system can now also recognize levels of noise and excitement levels of players, allowing it to remove bias when searching for highlights from players with a particularly vocal following or those who are particularly animated on court,” said IBM.
Earlier this year, Faultline Online Reporter was given a demo from the Watson team of some 3-minute round highlight reels applying AI to detect and analyze audio from crowds which are created automatically without even so much as a final human barrier for quality control. As we learned during this NAB discussion, IBM has been powering Wimbledon experiences for decades, although naturally Watson is a more recent foray for the London tennis club. Watson has also produced match highlights for The United States Tennis Association (USTA).
Watson Media divides its products across five core technologies – Captioning, Video Enrichment, Recommendations, Highlights and Streaming. IBM’s cloud infrastructure clout naturally helps sell these services, based on the important latter streaming segment which handles video hosting and transcoding as well as speech to text and analytics in one cloud video platform. Of course, these technologies and services are all exclusive to IBM Cloud.
Watson and Wimbledon also collaborated on what they are calling a progressive web app, purporting to cater for audiences in low bandwidth areas where generally less advanced hardware is available, so essentially a stripped back app presumably with streams limited to 720p and below, with fewer features available.
“Acknowledging that over 900 million fans in India for example, express an interest in Wimbledon, the new progressive web app is designed to provide a lightweight experience to ensure that fans in those territories can make sure they don’t miss out on the latest scores and results,” stated IBM.