AI and science with Demis Hassabis | The Royal Society x Nobel Prize

So, I suppose we might start by looking out at the audience and looking at the variety of people here and thinking that when they come to this room, they're not perhaps all thinking of AI in the same way. What would you say to that?

Yeah, that's a very interesting way to start the conversation because we produced this report. I have a copy here a couple of years ago, and at that time, even in two years, we've come a remarkably long way in terms of people understanding what AI is. I think if I asked everybody in the audience, AI means different things to them. Back two years ago, when I was asked how do I need to understand about AI, lots of people would say how do I learn about CNN and understand how it might work. Now you do not get asked that so much because people have been using things like ChatGPT and they understand some of the principles of large language models, but they're, of course, very different things. So we have this notion, including in science, that although the technology moves on and we have different capabilities, people's personal views of AI have dramatically evolved, maybe even faster than the AI has in some sense in the last couple of years. Going to the report, we set out the report was published to the month two years ago, and that meant over the 18 months before that we decided to do a piece of work on science and AI.

So, if I must just frame that you were chair of the working group, yes. Can I just ask you what the purpose of the report was? What were you setting out to do?

The purpose of the report came about from our science policy work deciding that we wanted to do a piece of work to understand how an area of disruptive technology was affecting scientists. One thing I said was do not come back and say AI. But of course what happened when you go out and talk to people, the most disruptive thing that either then or they could see coming was to think about how AI would affect particularly the scientific process and methodology as well as really what a scientist does. So we decided that that would be a really interesting thing to look at because although we hear about capabilities coming through, we hear less about how disruptive this is. I think what you would find in the report is that that's very much the type of thing that we were focusing on and also people are concerned about today.

And how did you do the work? You had a working group but then you solicited a far wider input.

Yes. It was talking to fellows. Some in the room there, over 100 fellows had input through attending either workshops we had. Workshops on things like access to data at the time, in the UK it was hosting the first AI safety meeting. So we had workshops on safety, large language models, and it was about trying to get a sense from these interactions to understand some of the issues to frame the piece of work.

And it's a big volume with lots of findings, but do you want to attempt to highlight the key topics that you thought were important?

The key things that I think are still also very relevant today: thinking about access to compute, some of that certainly in the UK has advanced, and the infrastructure needed to be able to work with any type of AI methodology. Access to data, I think that's one that's very much on people's mind, and access for your use is also an international issue. Reproducibility—this is not a new concept—but the challenges you have in being able to reproduce work and be responsible in the use of AI tools. Finally, how this is affecting skills and what people do was another aspect we were pulling out. AI often requires many different people to take on different roles. That means you're moving from siloed areas where an individual might work in their own office if you're a computational scientist, but now you're going to be working to do team science. Team science is very natural for certain areas of science, and other areas it's a brand new thing. Then it brings in things like how should you partner or should you compete? Scientists love to be competitive. So how do we manage some of these big cultural changes? These were all the things that bubbled to the top when we were thinking about the report.

There's an awful lot in there, and it's a beautiful framing for much of what we'll talk about later. You at the beginning of the report asked the question whether AI is assistant, peer, or tutor, and that again is something we will come to. But you came up with recommendations as well, and those were at a slightly different angle to the topics that you identified as most important to think about. Do you want to talk about those a little?

Well, I think that some of the recommendations were what I've said—access to data, being able to come up with standards for data. A lot of this is discipline specific; you can't have umbrella rules. Also, I think it's reflecting that areas that are new to using AI methods should be learning from those that are more advanced. The ones that are more advanced, we're going to hear about later, where you set things like data challenges. You've invested in having high quality data; those were ready to take on machine learning as being a tool that you could discover new findings.

As somebody who works deeply in this area already—you work on ultrasound imaging and on what you can get from multimodal data sets, so you're well into it—do you find that people are keen to learn from your experiences, or do you find that there are disciplinary boundaries to people's willingness to learn?

Well, in my own area, because I work in healthcare AI, as you said, I've done a lot of work in multimodal and ultrasound. What's been really interesting in terms of collaboration is that about a decade ago, it was very much computation—you were working asking for data and developing computational methods. It then went to co-design and co-creation of solutions with the end user, in this case a sonographer who is doing the scanning. And now we've moved on to as you deploy systems, you bring in experimental psychology to understand do people trust what you've used, how do they become reliant on the tools? Again, these are things that in certain areas are emerging, and we're going to hear a lot more about that in the future too.

The report is a production of a UK-based organization, but you are also Foreign Secretary of the Royal Society, so you travel a lot and have these discussions around the world. Do you find that attitudes to AI are broadly the same around the world?

No, they're not. It's been very interesting to have my role at this time when everybody wants to talk about science and AI because it's affecting people in different ways. I think there are some common things that we all would like to understand how science evolves—things like reproducibility, the effects of AI on publishing, peer review, writing papers, how the publishing of science will evolve. Then there are things like not everything translates to different parts of the world. In Europe, very much that is associated with responsible AI in science, underpinning everything. But if you go to India or China, they are much more excited by the potential, and you hear a lot less about the concerns, but they are also keen to talk about the ethical and responsible issues. I was at the AI summit in India, and things like the fact that large language models are built on Western world data is a big concern because it doesn't translate well into their own languages. So there are many things to have conversations about internationally. Whether this evening the conversation will get as far as international agreements and regulation, I don't know, but it seems of fundamental importance that you are having these conversations at every level to understand each other's perspectives as groundwork for future agreements. We find that as national academies is often a good place to have some of these conversations because we're talking about science that should cross geographical boundaries. The real barriers are things like access to data and being able to share information. The conversations for scientists in areas that affect global science are the ones we're having a lot of about at the current time.

Thank you very much. A couple more things and then we bring up our other panelists. One is that this was published in 2024, so a lot has happened and work continues at the Royal Society. Is the next iteration another report or how does it go on?

I don't know if head of policy is in the room. I don't believe the plan at the moment is to have another report of this size, but certainly looking at how machine learning has evolved. For example, we do not talk about agentic AI—we mention it, but two years ago it was not so prolific. Also, thinking about reasoning and uncertainty within the context of science; we need to be able to understand uncertainty, that might also be something we look at. But watch this space for future pieces of work.

And then this is a bit difficult, I think. Has the report had the impact you hoped it would?

Yes, in the sense that the purpose of a report is to start the next conversations and to provide a point of engagement to interact both nationally and internationally. From that perspective, it is a report that's had big impact.

Well thank you. It must have been an enormous amount of work, and listening to you now has set the scene beautifully for this evening's discussion. You've covered almost every topic raised in the questions from the audience, so I could just leave it to you now, but I'm going to invite others. Could I invite Paul Nurse and Demis Hassabis to join me on stage please?

Thank you both very much indeed. I don't think great introductions are necessary—you were introduced at the start. Demis, if we could start with you. You are CEO and co-founder of Google DeepMind, also CEO and co-founder of Isomorphic Labs, and you have used this phrase that what we're witnessing is science at digital speed. Perhaps just set the scene to bring us up to speed with what we've seen so far, what AI has done to science so far.

There's an awful lot in what you just said. The choice of the word 'foothills of the singularity' is an interesting one.

I want others to jump in, but may I just ask—when you talk about AI contributing to the discovery process, there were two papers published in Nature last week, one from Google DeepMind, where AI conducted experiments and made hypotheses. The pace of scientific discoveries filtering out into general use and acceptance is a strength of science, but it takes time to assess things. Do you feel people have that time anymore, or are things moving too fast?

Thank you. Alison, you wanted to jump in?

One of the challenges is everyone wanting to be first. We do not reward people for slowing down, particularly in academia—you won't get a paper published if you're not first. We need to think about reward systems that encourage people to explore and understand methodologies, which is important for going forward.

Thank you, very nice point. You two are both Fellows of the Royal Society, both Nobel laureates, and you have a long history together.

We do. Even dress the same.

I think we first spoke when you were an undergraduate at Cambridge or just after. We talked about a PhD, but you wanted to set up a gaming company, which you disapproved of.

I thought I can't handle that.

We haven't got very far, but we will do soon.

Okay, let's continue with this discussion of what's been happening or what will happen. What sort of problem is amenable to the approach you are spearheading? I'd like everyone's input. For instance, on Isomorphic Labs, which is a very ambitious project to reinvent drug discovery—you're an adviser, Paul, so that would be good to have your thoughts.

I wanted to go back a bit. I'm a wet scientist, so what has AI done for us? In my lab, machine learning now lets us analyze images in minutes instead of days. Literature surveys with LLMs get you started, though it's boring what you get back. AlphaFold has been amazing—we can test hypotheses quickly. The next step is applying a connected loop where AI helps at every stage of the scientific process. This requires integrating data from different sources, not just one type. Much biological data isn't collected in a uniform way, so we need methods to connect different types and lengths of data. The ultimate objective is understanding life, starting with the cell.

We'll come back to the virtual cell. Alison, did you want to chip in?

In my space, it's similar. You can't robotically perform an ultrasound scan—human expertise is required. We're thinking about solving individual tasks and integrating them. The ultimate would be a robot performing all tasks together. Five years ago, that was a dream; now people think it's possible.

Things are moving so fast. You're talking about AI and humans working together, but young scientists worry about their career paths and the role of the creative scientist. Let's explore that.

First, doing research in a wet lab is enormously boring. Most time is spent transferring small volumes of liquid. We should get robots to do that. At the Crick Institute, we have technical cores but not yet integrated as we'd like. But what we need to keep and develop is creative thinking. LLMs are not creative; they might push you in directions, but the human mind should be involved. The focus should be on how these tools assist humans to be more creative. What excites me is working with the machine to generate creative thought.

But it's true that while pipetting, that can be a good time to think about experiments.

In the 70s, I had time to think under the microscope. A problem with complex technology is you spend time making it work instead of thinking about the biological process. If it's simple, you have time to think. If robots do it, maybe we have enough time to think.

I'm just imagining a graduate student wanting to go do some deep thinking.

I'm just imagining my graduate student wanting to leave to do deep thinking.

There was a lot in what Demis said. Do you want to come in on the engineering question, Alison?