Machine intelligence researcher , of the , has been awarded funding by the Engineering and Physical Sciences Research Council () to conduct the research, which looks to leave a “legacy in AI development”.
As part of the project, he will work with the University's and Chelsea FC Academy to teach AI how to make split-second decisions and respond to others in a football setting.
Though sport is one of the focuses of the research, the decision-making skills are transferable, and the algorithms Dr De-Silva develops could benefit “a variety of high growth areas such as driverless vehicles, video gaming, and assistive robots”.
It is hoped the collaboration with Chelsea will also directly result in the creation of a new tool that can be used by football academies to assess players’ decision-making skills and answer ‘what if’ questions related to gameplay.
Teaching AI how to make decisions
If AI is to thrive in the real world, it needs to be equipped with cognitive – ‘thinking’ – skills such as those possessed by humans and able to quickly respond to difficult situations that arise in multi-user environments.
Currently, most AI systems are taught how to identify the best course of action using a technique called ‘reinforcement learning’ – a type of dynamic programming that trains AI by giving ‘rewards’ when it responds as desired.
AI needs to be equipped with 'thinking' skills if it is to thrive in the real-world. Image courtesy of Getty Images.
However, this type of optimisation and modelling is difficult to use in situations like driving on the road where multiple active objects – referred to as ‘agents’ – are involved.
This is because the best course of action is not as obvious in these ‘multi-agent environments’, so it is harder for AI systems to learn what to do, and this often results in inaccuracies in the ultimate decision-making processes.
Dr De-Silva’s research will focus on ‘imitation learning’, a technique that trains an AI system to choose a desired course of action by teaching it how human experts have acted well in similar situations.
This type of learning is common in humans as we too learn new skills by imitating those who have already mastered them.
Chelsea FC Academy collaboration
Football has been selected as one of the focuses of the research as the beautiful game is an example of a ‘multi-agent environment’, where ‘agents’ refers to players.
Players are required to make quick choices on the pitch and sense opportunities through anticipating what other players will do.
Dr De-Silva will work with Chelsea FC Academy to teach AI these transferable real-world skills using an extensive dataset on player and ball-tracking that has been gathered by the Academy over several years.
An extensive dataset on player and ball-tracking will be used to train AI. Image courtesy of Getty Images.
He will create an AI computer model of a footballer that is able to make informed decisions on the go, based on how real footballers – the ‘human experts’ – have responded to similar circumstances.
It is hoped the research will contribute towards the overall project goal of creating a system that can think like a human and act much faster than any person is able to, benefitting a variety of industries that use multi-agent systems.
The collaboration with Chelsea also looks to directly result in the creation of a tool that could be used by football academies to assess players’ decision-making skills and answer ‘what if’ questions related to gameplay.
Dr De-Silva says players’ skills could be measured against the computational model and coaching staff could also use it to visualise game strategies by seeing how the AI footballer responds to certain tactics.
He envisages the tool may also be used in sports broadcasting to give commentators a more realistic idea of what players may have been capable of in competitive matches.
It is hoped the project will also result in a tool that can answer 'what if' questions related to gameplay. Image courtesy of Getty Images.
Dr De-Silva commented: “I am delighted to be awarded this grant under the Artificial Intelligence Theme from EPSRC and partnered by Chelsea FC Academy.
“It is evidence of the timely need to bridge the gap between recent theoretical advances in AI and application-level challenges that need to be overcome.
“I believe this funding will enable me and my team to start to make a significant impact in the intelligent mobility and sports analytics domains while overcoming some key theoretical challenges in the area of data-driven policy learning in multi-agent systems.”
Chelsea FC Academy Head of Research and Innovation, Ben Smith, added: “We’re incredibly excited to support 天堂视频’s research in Artificial Intelligence. Whilst we hope to directly benefit from insights into player decision-making, we also hope to support the wider development of AI outside of sport.”
The second part of the EPSRC project will see Dr De-Silva teach AI in autonomous vehicles how best to tackle difficult scenarios on the road using data gathered on human drivers. More on that project theme can be found here.
Additional comments:
, Director of the Institute for Digital Technology, added: “We are very excited about the recent developments in AI applied to many applications.
“Our world-class expertise in multimodal signal processing coupled with advanced data analytics and modelling knowledge puts us on the forefront of this field and I am looking forward to seeing ground-breaking innovations and their applications to a variety of real-world applications in the very near future.”
, Director of the , said: “This is an exciting project that highlights the increasing importance of the use of technology in sport. Working with our partner the Chelsea Football Academy, Dr De-Silva’s research will break new grounds and provide unique and original insights into how AI can be used to develop and enhance the decision-making skills of elite footballers.”