Christian Eidem, the European based sports entrepreneur and Nils Lahr and Birger Steen, both Seattle based, and all three Norwegians, founded the AI company Orions Systems to develop machine learning capacity to analyse the data emerging from video systems.
Christian Eidem, co-founder, with International Sports Management and the World Series of Football. Nils Lahrs is a computer scientist who developed Windows Media Player and MP4 while at Microsoft. Birger Steen, ran Microsoft in Russia and is now is on the board of Nordea, the biggest bank in Scandinavia and Schibsted, the biggest media group in Norway. Their combined experience shaped the direction of Orions Systems but also informed one of the most successful applications of machine learning to date.
In the 1990s video had grown rapidly as a means of sensing and predicting danger. In the 2000s and beyond the volume of video data produced as part of the overall exponential growth in data was immense. More data has been produced in the last two years than in human history up to this point. Given the volume of data and the complexity of the situations that could now be viewed, Orions systems focussed on marking the data’s journey from the “edge” were it was generated to the “cloud” were it could be stored and analysed as smooth as possible. They also created a suite of systems that allowed the user to interface with that data effectively. They focused on the organisation and presentation of the data as it needed to be applied to the business or security need.
Some of the major problems of ML that need to be addressed for systems to move from drawing boards and business plans to real world business solutions involve the processing of the data, making it legible, accessible and removing bias from the selection. Human bias in the organisation of the data presented can destroy the advantage of the big data analytics by producing the wrong answer to the question inputted to the system. Many data scientists are good at the collection and not so good at the scaling, the presentation and linkage between data and business uses. That is why, according to Forbes, only 10% of Machine Learning prototypes make it to production. So a core problem of Machine Learning and AI is that the vastness of the data seduces the designers to focus on volume of collection. But what matters is the application of that data to a service, product or analytical line that is really needed.
Four things need to come together for the application of ML to make business sense. Firstly, you need a huge amount of data. Next, this data needs to be linked to a business in which performance and delivery can be improved by exploring that data. Third, you need to take out bias inherit in looking at the data. Fourth, you need an interface that can make the data work for the people delivering the goods, services or advice that the customers want to pay for.
Nils Lahr, while still CEO of Orions Systems went on to also found Synergy Sports with investment from Mark Cuban, owner of the Dallas Mavericks. Synergy use Orions Systems and applies them to the analysis of sports videos. This is a long way from the setting up a CCTV camera to watch a building for crime. But it has all the key features needed. Lahrs remains CTO at Synergy Sports and Christian Eidem and Birger Steen remain as advisers to Orions Systems
There is an immense volume of sport filmed globally on a daily basis. In 2017 in the US there were 134,000 hours of broadcast sport. Professional sport is a huge business. In 2018 the sports market in the US was worth an estimated $71.06 billion, this is projected to rise to $83.1billion by 2023. So getting it right in sports performance makes fairly pure economic sense. Selecting the plays and players by coaches is inherently a subjective occupation based on unconscious and perhaps also conscience bias. It you own a sports franchise you need to be able challenge that bias. But if you are coach or a player, or indeed a fan or an agent, you also need evidence on which to base or challenge a decision and perhaps also your own feelings about a player and their worth or lack thereof. And with systems like VAR in the UK premier league, also to question the decision of referees. Finally, you need the results instantly if you are using this for real time judgements and accessibly if you are using it for more leisurely performance assessment. All of these elements came together in the Orions Systems based platform created for Synergy Sports.
Having established Synergy and proven that the application of ML to sports was viable, Cuban and Lahrs sold their investment to Atrium Technology, for a significant pay day of a reported $90m. The deal will allow Atrium to take the technology global. Still using Orions Systems platform and technology.
According to Bloomberg, in the year leading up to the sale, the Orions Systems platform and technology used by Synergy had captured, tagged and analysed more than 60,000 basketball games and 15,000 baseball games. The application of Machine Learning to this area is going to grow and grow. It is the perfect case study of when, where and how AI works. But not sure all the founders, especially Christian Eidem and Birger Steen, will as happy with the outcome.