Next time you watch a football game or a rugby match on the TV pay close attention to the managers and their support teams. They are much more likely to be consulting tablet computing devices as they are to be jotting down notes with pen and paper.
The tablet is receiving data from GPS devices in the players’ clothing and giving vital information in relation to performance, fatigue and other important metrics. This helps the managers make decisions in relation to substitutions, tactical changes and so on.
"AI is certainly being applied in sport right now," says Prof Michael O'Neill, founding director of the UCD Natural Computing Research & Applications Group. "It is a part of the wider ecosystem being applied to problems. It is part of analytics and everything that goes around that. We see parts of AI being used in different ways. The classic is Moneyball for talent identification to build teams."
His team at UCD designs training programmes for teams of field sports athletes and has been working with two sports teams – an elite senior inter-county Gaelic football team in Ireland and a soccer team in the United States.
“We took data for the Gaelic football team and created a training plan for them,” says O’Neill. “We then use an AI algorithm to predict the gains it will help them make for the next season. We are producing training plans to ensure players receive optimum doses of training stress to allow them to achieve desired outputs during a match. We know what the match-day performance demands are and the algorithm produces a training plan to match those needs.”
The results were quite staggering. “We were able to produce a plan that would deliver a potential fivefold performance gain,” says O’Neill.
“Our approach is built on real-world knowledge and real-world constraints. The fivefold potential gain is achieved with a 20 per cent reduction in the training load. What’s driving the algorithm is not all about maximising output. The objective is to make the gain while minimising the risk of cumulative fatigue and injury.”
The objective is always to augment human performance and decision making. “Working with humans is the key to this stuff,” he points out. “If you have a machine doing it on its own it could risk the wellbeing of the humans involved.”
It can also be used to assist referees in various sports. “It can monitor and detect events the referee might not see during games. Event detection is a big thing. We are now starting to see the development of vision technology and vision processing and the use of deep learning image-processing algorithms to augment decision making. It will potentially make more information available to referees faster. But it is important that the human is still there to oversee it and exercise judgment. There is always complexity in sport. There are very few black and white situations such as if a ball has crossed the line.”
Event detection can also be used for player wellbeing by measuring the load on players during a game. “If they are taking heavy hits it can measure the g-force in tackles. It can also measure their gait. It can monitor those things and suggest when a player should be taken off.”
The legal profession
The day of the robo-lawyer may not have dawned just yet but it may not be that far away.
“Irish firms are engaging with soft AI rather than deep learning,” says Mason Hayes & Curran AI partner Brian McElligott. “Many of them are using process automation and analytics technology but I don’t know if there is anyone engaging with transformative AI. I don’t know of any law firm using deal-running platforms either.”
These are the AI-powered programmes which will automatically generate all the documentation required for a merger or acquisition transaction.
“Technology is used a lot in discovery,” he adds. “We have had digital discovery for some years and we are now seeing AI being added to that. Twenty years ago, we would have had people going through filing cabinets to find documents for discovery. But think about all the emails pinging through your inbox every day. The number of documents to go through now would be unmanageable.
“What tends to happen now is the parties agree to look at a number of categories and then search for a number of agreed keywords within them,” he continues. “That might get it down to around 10,000 documents for the page-turning exercise. But there’s no value in that for clients and they won’t be enamoured at having to pay large costs for no value. Clients are saying they don’t want to pay money for work with little or no value. They want the machine to do that. We did a pitch for a plc recently and they were asking about where AI would be driving value and reducing costs.”
Mason Hayes & Curran runs an award for innovation in legal expertise. A recent winner is a piece of software which can cut the property conveyancing process from six weeks to two in a bulk environment. “This is process automation rather than real AI but is an example of what technology is doing,” says McElligott. “Clients want to see commodity work treated like a commodity and they want to see technology being used to save time and costs.”
He believes there will be a lot more change in the next five years. “It will be transformative. A lot of clients have very strong in-house legal expertise. They only want to pay for very high-level expertise externally and legal firms won’t be able to charge for lower level work. Overall, we see AI as a good thing coming into the industry. It will take a lot more of the mundane work out of it. Legal professionals will be reviewing what the machine spits out. If you have 300 contracts with service providers to review during an M&A process, the quickest and most effective way to do that will be by using a machine.
“The client might only want you to review the top five with the machine reviewing the rest,”he adds. “It will be more than just reviewing for errors or omissions. AI will be able to spot anomalous patterns that would be missed by a human and alert the lawyer to them. This is something that most buyers would be interested in as it is truly adding value to a client. Deal platforms will mean that deals will happen more frequently. Clients will be expecting much quicker turnaround times.”
