How artificial intelligence slots into research

It was coding a Tetris-winning AI agent that flipped Dr Donghwan Shin’s focus from software engineering to AI. Now he’s investigating how AI can transform the research environment, and beyond.

A man with multicolour Tetris-style building blocks against a black background.

While Tetris isn’t a typical reason for a career change, it was the catalyst for drawing Dr Donghwan Shin’s research towards artificial intelligence (AI). 

“Reinforcement learning, a subfield of AI, is typically used to generate the best possible action based on a state. During a voluntary module in my PhD, I wanted to see if I could create an AI agent that was better than me at Tetris.

"Although I’m not the best Tetris player, I was surprised that I could create something that scored better than me. It was at this moment that I realised I wanted to explore AI within my research area.

"I had been focusing on software engineering at the time, but this encounter opened up many opportunities that led me to the work I’m doing today” he explains.

The impact of AI on research

Dr Donghwan Shin is a lecturer at the Department of Computer Science at the University of Sheffield.

He was recently awarded an EPSRC New Investigator grant for his research on the intersection between software engineering and AI on autonomous driving and is pushing the boundaries of human knowledge by utilising AI across his collaborative research landscape. 

The answers to the biggest questions in research lie in the data. AI uses self-learning algorithms to recognise patterns in massive data to reach conclusions more efficiently and effectively than a human would.

But how will this shape the research environment?

“AI is like electricity or the internet. It is a huge, adaptable general purpose tool that we can do so many things with.

"The impact of AI on the research landscape is already evident through elements such as speech and image recognition, but the true impact it will have on the research landscape is tremendous” he adds.

“The use of AI in my research on autonomous driving testing is twofold. The autonomous driving system itself is a composition of different elements developed through deep learning and reinforcement learning.

"This combination creates a cutting-edge system that is constantly evolving. The system is also tested using AI approaches. I use machine learning techniques to generate complex driving scenarios, such as a car moving out of a lane, to predict the outcomes of a safety violation.

"Without AI, I would have previously had to do this manually, over and over” he adds.

A new Centre for Machine Intelligence at Sheffield

To harness AI’s potential to transform research across every discipline, The University of Sheffield has made a multimillion pound strategic investment into the new Centre for Machine Intelligence

Under the Turing Network Development Award from the Alan Turing Institute, considerable momentum has developed around AI at Sheffield involving over 200 researchers from over 30 departments and institutes. 

“Our goal is to create a better future through AI-driven research, innovation and education.

"To do this, we’re establishing an interdisciplinary research environment that harnesses the strengths of our University in manufacturing, autonomous systems, health, energy, digital humanities, and digital twins.

"This collaborative environment will enable researchers to expand their networks, access expertise from various fields, and elevate the impact of their work” says Donghwan. 

Shaping the future of AI

While AI is quickly on the rise, the full extent of its fruitful impact on the research environment is still unknown.

“We’re already seeing a flood of new research papers about AI every week. However, as with all research, it is important to conform to the standard legal and ethical norms.

"As researchers, we have a tendency to focus on a specific problem and ignore other aspects surrounding it. This can be problematic in terms of fairness, transparency and responsibility when creating AI-based solutions to real-world problems.

"While AI might help us reach a solution faster, we need to evaluate whether it is fair and responsible in terms of society.

"I believe this is one of the most complex, yet important, challenges the Centre for Machine Intelligence will try to address” says Donghwan.

“The pervasive nature of AI will lead it to becoming an integral part of our society. It won’t merely be a technology we use, but a close companion that we interact with extensively.

"It will redefine how we live and how we work, and entities such as governments, corporations and universities should cooperate to create laws and guidelines for responsible AI” he adds.

Written by Alina Moore, Research Communications Coordinator