AI, Academia and Business

Gosh, that's a lot of story. Well, at medical school I was captivated by the brain, and I did what would now be called an intercalated BSc in what would now be called neuroscience. I was convinced that that's what I wanted to do — but I didn't like neurology at all, so that was a bit of a problem. So I spent a lot of my early clinical training doing intensive care medicine, and became slightly convinced that I wanted to be a renal physician. I'm not sure why.

Fortunately, the renal physicians had identified that I was completely unsuitable for such a career and didn't appoint me to any of those jobs. So I had to go and become a neurology SHO, where I discovered that actually I'd been labouring under the misapprehension that neurology was boring. In fact it was fascinating, and then my neurology and neuroscience interests married up. That was really, for me, the start of a career in clinical academic practice — using brain imaging to image and try to understand consciousness: the set of subjective feelings we have about the world around us.

I think that's fascinating, because it's one of the most important things that gives our life meaning. Not many of us would want to live out the rest of our life unconscious — either in the sense of being asleep, or in the sense of being a zombie without inner thoughts and feelings. So it really is critical to human existence. I did a PhD, some postdoctoral training, and to cut a long story short became a professor looking at those areas. That then started a second phase in my career, where I discovered the joys of leadership. I was lucky enough to direct an Institute of Cognitive Neuroscience at UCL, and now I'm Dean of Life Sciences and Pro-Vice-Provost for Artificial Intelligence at UCL.

What's the neural basis of consciousness? I would hope to get to the bottom of that.

Yeah. So we use the term consciousness in at least two different ways. One is to refer to the fact that we can sometimes be awake and having experiences, and sometimes be asleep or unconscious. That's one set of brain mechanisms — what wakes us up in the morning, what puts us to bed at night, what can knock us out if we're hit or in an accident. I'm less interested in those.

What I'm really interested in is when we are awake — and of course sometimes when we're asleep, when we're dreaming — we have this vivid set of experiences of the world outside and of our inner self, which we call consciousness. We can say we're conscious of an apple, or a scent, or a flavour or smell, as well as more complicated things: I'm aware I feel a little self-conscious, or I'm aware I feel a sense of déjà vu. So we humans have this complicated set of inner experiences that we can share with other people through language and non-verbal communication.

We're not quite sure whether other mammals, other animals, have those experiences or not. Some of us might think our dog or our cat is conscious in the same way we are; others would not. Some people would think an earthworm is conscious; others would not. And the reason we have that ambiguity is because we don't really understand the neural mechanisms of consciousness — so we don't have any scientific basis to appreciate what animals might be conscious and what might not. Of course, neurological illness, but also some psychiatric and other illnesses, can affect our consciousness in all sorts of ways. So I'm interested in understanding that, which seems to be something special about being human and something that gives richness to all our lives.

It came about because the university had a realisation. Artificial intelligence is obviously quite a big term, but in the broadest sense it encompasses a lot of what you might call analytics — algorithms, ways of interrogating and learning from data to generate behaviour or decisions that seem intelligent. And the realisation was that this sort of activity was taking place all across the university — not just in traditional places where you might expect AI research, like a computer science department, but also in parts of our Institute of Education, in our Faculty of Laws where people were interested in regulation, and lots in medicine — some of whom I'm sure you've talked to or will talk to, like my friend Pearse Keane.

So all across the university there was this ferment of activity, but underpinning it we didn't have any sense of a narrative or a strategy for how, as a university, we wanted to promote and bring those things together so the whole was more than the sum of its parts. We also didn't have a strategy for what sort of computational equipment or software stacks we needed to provide for these researchers and teachers. And we didn't have any systematic way of engaging with the companies and social enterprises out there who are also using AI to work on real-world problems. So the post was created to think about those things, to coordinate them, and to try to make us — as a modern global research university — better at having an integrated whole that was more than the sum of its individually impressive parts.

One interesting area is whether AI has any role to play in therapies. We often think of AI in surveilling populations, in preventative medicine — predicting what you might have to do to avoid a particular disease state — or in the diagnostics you mentioned: could we use AI to read mammograms, or deliver insights from other medical images? Those are all in the diagnostic and "before" space. It's harder to think about how AI might contribute in the therapeutic space, because therapeutics are often — not always — medicines we take that alter our bodily physiology.

So one interesting, neglected area that a few companies are working on is the use of natural-language algorithms to derive talking therapies in mental health — cognitive behavioural therapy, delivered through different channels, and ultimately perhaps algorithmically, or through an artificial intelligence. That's the kind of area people don't always think of when they think of applications of AI. We might also think about how AI can work alongside existing therapeutics — whether medications or surgical therapies — to synergise or enhance their effects.

A final area is healthcare delivery itself. These days, certainly in Western settings, care is delivered in very complex, large systems of primary and secondary care, all integrated into what we might call a health science network. These structures are incredibly complicated to operate and have huge operational inefficiencies — and we see the stresses hospital systems can be put under in things like the current global COVID-19 pandemic. What we perhaps haven't paid enough attention to is whether AI and analytic systems have the potential to really optimise those aspects that aren't so medical, but are to do with how a hospital or healthcare system operates: the flows of patients, the throughput of an outpatient clinic, the timing of appointments — things that collectively make the patient journey much easier but aren't directly diagnosis or therapy. So that's a second set of areas where I think AI has the potential to have a major impact, but where we perhaps haven't fully harnessed the opportunities.

There's one paper — a colleague of mine, Amy Nelson, is the author — that I think you might be talking about, which was about trying to predict non-attendance at hospital appointments. The issue is straightforward: many people fail to attend their hospital appointments, and that creates — particularly if you've got very expensive machinery like an MRI scanner — essentially wasted resource that could have reduced a waiting list or allowed a patient to access something earlier.

So the very simple question she asked was: can you actually predict who is not going to attend their appointment? Because that also allows you to target what you do. Rather than ringing everyone up or texting everyone to say "remember your appointment," you could target your intervention — ring up the people predicted to miss and ask why. Do you need transport? Do you need some help? So precision medicine, if you like, but directed at appointments.

What she found was that, surprisingly, the routine hospital data held in every hospital about people's appointments can do a pretty good job of narrowing down who's most at risk of not turning up. That's a really practical thing, and it would save countless billions of NHS money if we could roll that kind of solution out across the entire health service. One final point: figuring out that this is a problem is something healthcare professionals can do, because we're in hospitals every day and we see people don't turn up. You might not see that from outside if you're not working in a hospital. So there's a real role for all of us working in healthcare settings to use that knowledge to identify those little areas that seem trivial but cumulatively are a huge wasted resource — capture them, and then use AI and advanced analytics to solve them and make the health service better.

No, I wouldn't necessarily agree with that characterisation. First of all, your idea that the rich benefactor has gone away — that's not true. There are amazing philanthropists out there who can and do invest considerable time and effort in supporting biomedical research. We know some of them because they're billionaires, like Bill Gates and the Bill and Melinda Gates Foundation, one of the world's largest charities. But there are also countless other philanthropists making smaller but really important gifts.

Similarly, it's not really the case that companies have just come along in the last five minutes and are suddenly supporting research. There's a rich history of companies either doing research themselves — a lot of R&D in this country takes place in pharmaceutical companies, technology companies and so on — or supporting that interaction with universities. So I'd see it more like an ecosystem, where there are different types of funding and different types of funder, and different situations in which research can take place. And yes, that does shift over time, just like any ecosystem. Think of a coral reef — although maybe the analogy is wrong, because there's no equivalent of bleaching and climate change here. That ecosystem is continually changing. What we have to be attentive to is thinking about what balance would look like. It would be wrong to have a research system solely dependent on government funding, but equally wrong to have one solely dependent on philanthropic individuals, or just on companies. So the more sophisticated question is: what's the right balance?

It clearly varies by the type of company. A company involved in retail or hospitality is perhaps less likely to be engaged in as much research as one delivering new medicines or creating new technologies for public benefit. So it differs across sectors — that's called R&D intensity, and governments measure it as a percentage of turnover.

The longer answer is that companies that do invest in R&D do some of it in-house — they have teams to develop their own ideas, which in the medical space might be surgical robots, or new agents to treat diabetes, or something completely different. But they also work with partners like universities and hospitals: sometimes because there's particular expertise they don't have in-house and want to access through partnership, sometimes because there's a need to work alongside patients and have the patient voice represented, sometimes for other reasons.

So it would be a mistake to see companies as somehow an optional part of this ecosystem — as if everything could take place in a beautiful world without them. We should remember that creating companies that make things is ultimately how we translate most benefits into the world. If we want to create a new medicine in a university or hospital, we can do so. But if you want to scale up the manufacture so we can make millions of doses and transport them all over the world, we're going to need a company — because we need to understand bioprocess engineering, scale-up of drug manufacture, logistics, regulation, all sorts of things where the expertise wouldn't necessarily be in a hospital or university. So it's not as simple as people sometimes see it in the health sector — that companies are somehow intrinsically less good and healthcare systems intrinsically more good. I think both are required to deliver healthcare advances.

Got to enjoy yourself. It's easier said than done — sometimes it's very hard to enjoy yourself when you've got a grant deadline, or you've just failed to get a job. But pulling back to the broader picture: if you're not smiling, or at least don't have the potential to smile, when you get up and go to work — if you're not enjoying what you do — then I think you need to take a long, hard look at it and ask whether it's temporary, just because you're having a bad day, or something more permanent.

The second thing for me is that I've always been afraid I would lose self-awareness — lose the ability to look inside myself and ask, "am I really happy?" I think doctors in particular, people with medical training, are sometimes really good at convincing themselves that whatever they're doing at the moment is the right thing. So what I've always sought to cultivate is a little bit of scepticism — a voice that prods me and says, "are you really sure? Are you really sure?" And then I go, "yep, I'm really sure," and I'm back to enjoying myself.

The third thing is work-life balance — remembering that balance is a verb, not a noun. It's not something you achieve when you're 22 and then say, "ah, got work-life balance, sorted." Everyone's life is always changing, everyone's circumstances are always different, and even when you think you've achieved stability, something out there — like COVID-19 — will come along, good as well as bad, to disrupt that balance. So adopting a continuous checking — "have I got the balance right? Do I need to change anything?" — has always been important to me. Those are three things — perhaps not all of them, but certainly three that are really important to think about.

"Too hard" is a funny thing to say, isn't it? It kind of depends on the person. Some people like to work like crazy, and that's really important to them — often they're really, really good at what they do, and they're surrounded by people in their personal life who are content with that. I think that's fine. Equally, I see people who aren't working like crazy — not in the sense of being lazy, I don't think that's fine — but people for whom going to work is not the be-all and end-all of their entire life. Maybe they have something they're passionate about outside work, or family members who are more important to them than work. That's fine too.

The important thing is that you're aware of where that balance lies for you. What's unfortunate is when people feel pressured to work too hard for reasons that aren't legitimate. Sometimes we all have to work hard at what we're doing — but if we're working really, really hard when we don't need to, that's a mismatch. And you can have a mismatch the other way, of course: you can be a bit lazy, not putting in the effort commensurate with the goal you've set your heart on. So the important thing isn't whether people are working too hard or not hard enough — it's whether there's a mismatch between how hard they're working and how hard they need to work to achieve the goals they want.

That many people want to be like me? I have no idea. But all right, let me say a few things. The first is that medics are often the first to forget the difference between a prospective and a retrospective trial. They think, "okay, we'll pick your favourite older or more senior person — what pearls of wisdom have you got about how to get where you are today?" But the person you're talking to is someone who's made a whole load of choices during their life, and you're assessing them retrospectively, at a point where I'm in the position I'm in. What we don't know is: if a thousand Geraints started off aged 20 and made exactly the same choices as me, maybe 999 of them would be content, successful and happy, able to talk to you on a podcast — but equally it could be that 999 of them are really miserable, their choices have led them to rack and ruin, and you just happen to be talking to the one who's really happy. So you've always got to take that kind of advice from your seniors with a pinch of salt, because it's not a prospective trial. We don't actually know if the choices I made are the right ones for you, or me, or anyone else. Now, that said — put me back on track. What do you want me to talk about?

Yeah. Two things — maybe three. The first is that I agree with the basic premise: it's really hard to see people in any area of life who are successful without working reasonably hard to get that success. Of course, there are always a few incredibly lucky people — lottery winners who write down some numbers and suddenly have millions, although that doesn't necessarily bring happiness. So yes, there's an element of luck, but most people work hard if they're going to be good at something.

The second thing is that it's about working smart as well as working hard. Just putting in the hours — you can either cut down on those hours by working smarter and have more time for other things, or work smarter and put the same hours in and be even more successful. But it's certainly not true that there's a linear relationship — that if you work double the hours you get double the success. So a little thought about that: it's about commitment and passion and working hard, but not necessarily going completely crazy.

The final thought is that most people work harder and longer at things they enjoy. So it goes back to finding what you love, rather than what you think you're going to be successful at, or what you think is the most important or sexiest or most urgent area. Finding what you love is probably more important, because then not only will you work harder, but you'll actually enjoy the hours you spend working — your work is a reward in itself, rather than a means to an end like earning a lot of money or being seen by your peers as highly successful.

And perhaps a final final thought — I know I keep saying that. As you get older you do reflect on your own mortality; certainly having children did that for me. Someone once said to me something I think is very true: nobody ever laid on their deathbed and said, "do you know what my one regret is? I should have spent much more time at work." No one ever says, "my one regret is I spent too much time with my family and friends." So in the final analysis, you've got to balance your commitment to working hard with the other aspects — which will be different for all of us — that make your life rich and enjoyable, and make sure you've got plenty of time for those things. Easier said than done, I know, and it's a constant balance. But it's a really important habit to get into, and to keep tinkering with, to see what's right for you.

You should come downstairs from where I'm doing this podcast — my house is full of books. Books are just the most amazing things. The invention of writing, and the ability to put knowledge outside one's brain and pass it on to people who aren't in your immediate social circle, is one of the most amazing inventions in the whole history of humanity. Books have been all around me since I was little, and now they're all around my kids — piles of books everywhere. I have one of those piles — there's a Japanese word for it — of books on your nightstand that you vaguely regret buying and not reading, and feel slightly guilty about.

Are there single books? That's an almost impossible question to answer, unless you're about to invite me onto Desert Island Discs. I still read quite a lot of fiction — not much science fiction, but a lot of fiction. For example, I've just finished, this morning, a David Mitchell book — a contemporary author who's just got a new book out that I'm going to read on my holidays, called Utopia Avenue. He writes wonderfully. The one I just finished is a fiction about a Dutchman in a trading outpost in Japan in the 19th century, which sounds very unpromising, but it's a very rich story of unconsummated love, political intrigue and 18th- and 19th-century seafaring — a fantastic book. The point, for me, is that fiction lets me go outside my daily life, outside the job and the work I do, and escape into a different world. I've always admired authors who can create those kinds of environments and worlds for you to inhabit almost out of nothing — just sitting on a chair, you can be transported away.

I'd recommend not reading the books, actually — I'd go straight to the journals. Yes, you can go to Amazon and find any number of good introductory books or textbooks on consciousness or cognitive neuroscience. But I think a better way is to learn how to navigate the literature through PubMed and publicly available resources. Now, you've got to have access to a subscription or a way of getting hold of the PDFs — but following your interests, starting with a review article from one of the major journals, a Trends journal or a Nature journal, something you find interesting, and then following the chain through the references, out into "oh, I'll have a look at that one, oh that looks interesting" — I still do a lot of that myself. I spot a lot on Twitter, and as Dean of Life Sciences I get a lot of stuff, so I'll find myself reading about the energetics of bird migration and think, "wow, that's really interesting." Just allowing your imagination and interest to wander a bit more freely, rather than saying "now I'm going to read a textbook on X," is a better way to stimulate your interest.

So if you're interested in AI and technology, stick it into Google — make sure you go to Google Scholar or something that's actually indexing peer-reviewed scholarly work — and off you go: loads of preprints, loads of articles. Don't spend too much time on each; it's not an exercise in learning, it's an exercise in saying "this is really interesting, so I'm going to read it," or "this looked interesting, but as soon as I read the first sentence it was very boring, so I'll chuck it away and go on to the next."

Of course, nothing beats a good bookshop. Rather than read a book, I'd say get on a train, go to your nearest large academic bookshop, and just wander around the sections that interest you — pull random books off the shelf, have a look, and see what interests you. You won't be able to do that in your local Waterstones if it doesn't have a neuroscience section — it'll be popular science — so you might have to take a little trip to your nearest large university bookshop. But that's a rewarding thing in itself: the joy of browsing and picking up something that interests you. It's a habit I got into many years ago, when I used to go and see my grandparents in Cardiff. Every year we'd go to a bookshop — I'm not sure if it's still there — and the rule was I was allowed to buy one book. Any book at all. So I wandered freely, and for a period in my teenage years I was obsessed with armoured fighting vehicles — encyclopedic books of Soviet tanks and NATO tanks and things like that, which must have horrified my poor old grandparents. But no questions asked, they always bought a book, whether it was cheap or expensive — one book from every visit. That love of books, reading and knowledge, which they and my parents helped impart to me, has stayed with me to this day.