Data Capital: University of Edinburgh’s School of Informatics introduces GAIL force wind of rapid AI change
When Helen Hastie was an undergraduate studying linguistics and artificial intelligence (AI) at the University of Edinburgh in the early 1990s, it was a very different era in technological terms.
“Back then, the computing power of your phone would fill a whole room, and now we have the world’s fastest computer coming to the University,” she observes.
Hastie is now at the forefront of rapid technological change at the University – as head of the School of Informatics, which hosts the new Generative Artificial Intelligence Laboratory (GAIL).
GAIL is the latest initiative in the 60-year journey of artificial intelligence at the University, and as AI becomes increasingly vital in all areas of our society and economy, it’s an exciting time for Hastie.
With her colleague and natural language processing expert Professor Mirella Lapata at the helm as acting director, Hastie is particularly energised that GAIL can reach across the whole University.
“It’s based in the School of Informatics because we have the critical mass of AI and machine learning experts, combined with natural language processing and biomedical AI and other areas, but it’s a University-wide initiative,” she explains.
“The key to success is joining up with the rest of this world-class University – medicine, engineering, business, law, geosciences, to name just a few disciplines.
“The University of Edinburgh has been researching AI for 60 years, and we need to be at the forefront constantly. Generative AI is usable for everyone – from someone sitting at a computer at home, to small businesses and large companies. Everybody can benefit.”
When the University launched GAIL, Hastie talked about generative AI “curing diseases, improving living standards, protecting our environment”. But what might this look like in practice?
Curing diseases
“Generative AI can process massive amounts of data, discovering trends across many different experiments very quickly. We believe it can make the drug discovery process more efficient, reducing the development cycle which can last more than 12 years to far fewer and improving drug safety. It can also help with diagnosis of heart attacks, cancers and early detection of UTIs [Urinary Tract Infections, common but hard to diagnose with certainty in older people], and support healthy ageing.”
Improving living standards
“It can help the economy by increasing productivity. If a worker can automate part of their job, it makes them more efficient, and companies more profitable, improves quality of life and makes jobs more interesting. We need to understand what those future skills are and build a new skills pipeline.”
Protecting the environment
“Generative AI can provide valuable insights into environmental trends, making things more transparent and enabling better decision making. It’s all very well, somebody sitting analysing all this data, but you need to take it to policymakers and explain it simply – generative AI is good at doing that.”
Hastie also wants to utilise the wider expertise of the University to address the complexities and challenges of generative AI. “Thinking about responsible research and ethical design is vital from day one, not as an afterthought,” she maintains.
“We want AI to be safe, effective, and trustworthy. Everybody at the University is aware of that, and we are doing this through our Centre for Technomoral Futures.” [part of the Edinburgh Futures Institute and led by Professor Shannon Vallor, a globally recognised expert in technology ethics].
Helen Hastie’s background is in human-robot interaction, so where does robotics fit into this future GAIL vision? “Robots traditionally have been in manufacturing where they do repetitive tasks,” she explains. “We want robots to be intelligent and come out of the closed environment into the real world, which is very dynamic, unpredictable and constantly changing.
“Humans are very good at adapting quickly and understanding what’s going on and what’s safe to do physically. For robots to have that element of natural intelligence, they need sensors. And when you put sensors on a robot, you get a lot of data very quickly. So one aspect of generative AI is to take all this data and respond appropriately, for example, by following or giving instructions through speech, language, and vision. I think that’s absolutely vital.”
During our interview, Hastie regularly uses three adjectives about AI – it must be trustworthy, accurate and transparent. “We need to have the confidence that the technology will work and – if it does break down – to be able to understand and explain what happened and move forward,” she says.
To be able to do this, and fulfil the potential of generative AI, requires a trinity of experts in machine learning, data science, and high-performance computing.
Hastie says that the Capital is very well placed in this respect: “It’s so exciting to be in Edinburgh, which is soon to host the UK’s first exascale computer – one of just a few in the world.”
But what about the huge amounts of energy and water consumed by high-performance computing and AI, an issue highlighted at The Scotsman’s Data Conference in September?
“It takes massive amounts of CO2 to train large language models such as ChatGPT and Bard.” Hastie explains. “We’re aware of this, and one of the goals of GAIL is to explore how we can make large language models smaller and more environmentally friendly.”
The picture painted by Hastie of GAIL is an initiative that covers the whole University and which seeks to address an enormous range of societal, economic and environmental issues in a safe, responsible and transparent way. Is she intimated at all by the sheer scale of the job?
“It’s really complex and hard, but it’s also really exciting,” she insists. “We have the opportunity to bring world-class experts together across all kinds of fields – such as linking AI and genomics and using AI to save lives. Those connections are where the real power lies.”
And does she feel the weight of history on her shoulders as one of the latest carriers of the University’s baton of 60 years of AI excellence?
“We have this great legacy but don’t rest on our laurels,” she says. “We’re top-30 in the world for computer science, and we are home to the largest number of informatics academic researchers in the UK. We have to understand the latest techniques and create new approaches to tackle today’s challenges.”
Questions and Answers
What exactly is generative AI?
“Generative AI is the class of AI algorithms and machine learning systems that use large quantities of data to generate things. People are most aware of text, but it can also be images, audio, video, computer code – and even genes and molecules or proteins.”
How do you decide where to focus when AI is everywhere?
“We do research discovery from the ground up, but we also try to understand the key challenges in society and the economy. At the University, we work with industry and public sector bodies on translating core research to have a real impact on society.
“If we have powerful computers, a strong data set, and a problem, that’s where GAIL can be really powerful and progress really quickly. You need to have the challenge and really good data – then your experts can develop methods to make the models more accurate.”
How can you help shift the debate from the existential risks of AI to more practical uses?
“The benefits are huge and it’s a question of educating people about the positive difference it can make to problem solving, scientific research, creativity, and security. But we – as scientists – need to work closely with policymakers on governance and regulation so we don’t stifle innovation, but ensure what we do is safe."
Where does GAIL fit into the broader global AI picture?
“There is a lot of noise around AI at the moment, which can be unhelpful to industry and policymakers who know AI is important but are unsure how to best use it. What GAIL offers is a fully-rounded approach to tackling the problems they are trying to solve with AI. We have access to some of the most advanced computing power on the planet, twinned with expertise in using data for good. Because GAIL is based at the University of Edinburgh, computer scientists are working alongside the likes of sociologists, legal experts, ethicists and artists to solve key challenges, including working with the Centre for Technomoral Futures. I think this really sets us apart at a global level.”