Scroll to top

LAB CHAT with Peter Riley

Our new “Lab Chat” guest is Peter Riley. A Consultant at Decision Lab during the working week, he is a keen trombone player who does not shy away from his fans. PlayStation had a role to play in defining his career choice to become a simulation expert.

In our chat he touches on the hot topics that worry the industry, such as: why should you trust a robot/AI making tactical decisions in football? And what are the untapped yet promising areas for simulation to explore?

What do you enjoy most about your work?

The thing I enjoy most is tackling complex technical challenges. Each client has their own unique requirements, and I find it really interesting to design the architecture of a simulation in order to meet their specific needs. During a project’s discovery phase, we discuss with the client exactly how each piece of necessary functionality will be implemented. There’s a great deal of autonomy when working for Decision Lab, and I thrive when given the freedom to test and apply my own approaches. But alongside the autonomy, there’s a highly-skilled team of people who are there to support when you need it; I think we all continue to stretch and improve each other.

How did you come into this profession?

Throughout my A-levels and degree in maths, I was most interested in the practical applications. In particular, I enjoyed the operational research kind of modules containing things like linear programming, routing algorithms, and system dynamics. But before all that, my first exposure to simulation was as a child when I played the FIFA games on my PlayStation; in the career mode, you pick a football team to control whilst all the actions of the other teams are simulated for decades into the future. I found it really interesting that the game was predicting all these events like who would win the league and who would become the best players, but I didn’t have any idea that there are careers in building these simulations.

The best bit about working at Decision Lab is…

In general, there’s a lot of flexibility within Decision Lab. It’s nice to have flexible working hours and the option to work remotely whenever we choose. Moreover, the company is versatile in the range of clients that we work with, and in the approaches that we use to support clients. This leads to a lot a variation between projects, which is great for learning a breadth of skills and software, and it’s refreshing to continue to face new challenges.

One of the most interesting projects was the Cyber Arms Race Experimentation (CARE), which looked to develop a capability to predict and explain strategies that might be employed by cyber- attackers and cyber-defenders, and provide recommendations for countering them; the client for this project was Dstl. The core technical component is complex: attack and defence agents carry out actions in a simulation that is a representative military environment, in such a way that action strategies can be developed via Deep Reinforcement Learning. I enjoyed the collaboration we did on this project with our technical partners: Actica Consulting (who have in-depth subject matter expertise) and DIEM Analytics (who have developed a Framework for Automated Strategy Extraction, FASE). It’s interesting when a project combines such a range of approaches, and everyone can learn something from the other specialists in the team.

What’s the future for simulation and AI?

I think there are still big opportunities to develop the use of simulations, and more broadly AI, in sports industries. Many sports have been slow to adopt the use of data to inform their decision-making. There’s a common criticism of AI which I think is particularly prevalent in sports industries: why should you trust the intelligence of a machine and why would it be better than the decision-making of a human with many years of experience? No matter how good AI gets, it’s difficult to imagine a robot on the touchline, deciding on tactics like formations and substitutions. To improve trust and thus become fully accepted, AI systems must be interpretable and reveal uncertainty awareness. As previously mentioned, we’ve had an explainability model analysing the behaviour and strategies of DRL agents; I anticipate that the demand for these techniques will continue to grow. However, there are certainly other aspects for which an increasing number of sports teams are already utilising the power of AI, such as: predicting when players need a rest to avoid injury, and identifying new players to buy who would make the most valuable improvements to the team. I’m sure we’ll be seeing many more use cases in the near future.

What’s the post COVID work and life look like?

Like almost everyone else, I’m really looking forward to being able to meet with people face-to-face again. Even after mastering the art of video calls – unmuting myself at the right time and ensuring that there’s no dirty washing in the background – there are still occasions where working with my team in the office would be a bit more efficient. Plus, I’ve really been missing the free fruit and endless supply of biscuits in the office.

Personally, it will be good to get back to playing with my brass band again. Playing at some of the London park bandstands was becoming a regular thing each summer, which we missed out on this year, but fingers crossed we’ll be back in 2021. I’ve had one request to play my trombone in a video catch-up with my work team; I politely declined but I’m always flattered to receive attention from my adoring fanbase.

Author avatar
Decision Lab

Post a comment

We use cookies to give you the best experience.