Exclusive Interview with Nobel Prize Winner John Jumper: AI's Next Frontier After AlphaFold

Recorded at the ISMB/ECCB conference in Liverpool, this episode features John Jumper, who led the Google DeepMind team behind AlphaFold and shared the 2024 Nobel Prize in Chemistry. He talks about how AlphaFold predicts protein structures, why the Protein Data Bank was essential to it, how scientists came to trust its predictions, and his interest in getting language models to reason about biology. Laila Rizvi is joined by co-host Stephen Horn and structural biologist Luke Yates.

Watch on YouTube

Transcript

Dr. John Jumper led the team that built AlphaFold, an AI model that reshaped molecular biology by solving a decades old challenge. He was awarded the Nobel Prize in Chemistry in 2024. Now, as a distinguished scientist at Google DeepMind, he's focused on the next frontier using AI not just to predict biology, but to design it.

We're in Liverpool at ISMB/ECCB where the cutting edge of computational biology meets the future of AI. Here, the talk isn't just about solving problems. It's about reimagining the questions we ask. I'm Laila Rizvi and I'm joined by Stephen Horn, my co-host, and Luke Yates, a guest of the podcast. Guys, it's great to have you here again.

It's great to be here. It's great to be here in Liverpool.

So AlphaFold, John Jumper, big discussion. Amazing guest. You know, obviously a real trailblazer, very instrumental in the AlphaFold project of course. And really fascinating keynote lecture last night, was really rather amazing.

So this is something you know a lot about, right? So, so what's the one question you want to put to him? You've probably been thinking of this last night where you were thinking, what am I going to hit him with? So what's the one question you want to ask him?

Well, I think as a structural biologist, I'm particularly proud of the community. The fact that we generated a very large open access repository of protein structures, which are experimentally determined with jealously guarded, that in terms of, you know, making sure the structures deposited reach certain criteria. And of course, that was instrumental as a training model for AI. For, for AlphaFold in particular, for generating these predictions from proteins, you know, sequences. And, and I guess I kind of want to know how instrumental is this type of data for building large language models.

Luke, can you just give us a little bit of a overview of what AlphaFold is and why it was so monumental?

So in, in my view, solving protein structures is really essential to understand how life works. Proteins are the building blocks of life. They do the work in our cells, in our bodies, to help to keep us alive, and, and we have a huge amount of genetic data. And we know all the sequences of lots of proteins across all kingdoms of life. Right the way through to microbes, you know, at the bottom of the ocean.

But one of the challenges is, is generating that structure requires months and years of dedicated work in a laboratory. And so there is a, an age old problem that the challenge that AlphaFold solved was, can we predict accurately the three dimensional structure based only on the amino acid sequence? And the answer is yes, quite accurately. And AlphaFold was that breakthrough that allowed us to, to uncover and to be able to do that kind of work at a fraction, in a fraction of the time. And so we don't need to do that laborious work. In the laboratory, we can, we can generate the structures on a computer.

And Stephen, what's so interesting to you about this area? I mean, I know you're not a scientist like Luke, but this is Google DeepMind. They made huge strides in a field that they weren't even involved in and they wiped the floor clean. So what do you think of that?

What do I think of that? Well, I think it's all about innovation, isn't it? And I think it's about that sort of focus and the sort of drive that, that, that, that brings to fields, which are academic fields, if you like, which is just fascinating how you can, in that awful American word, monetize. Right. Sort of science and technology without trying...

So I really fascinated to see, I'm also fascinated to see a prominent scientist in industry. You know, we've been doing this for 20 years, and most companies struggle to get prominent scientists to work for them because they want to work in what's known as public science. So be really interested to see how John, Nobel Prize winner, deals with that.

Sure. Yeah. And I, I'm also interested in zooming out a little bit. And obviously AlphaFold came about quite some time ago compared to the ChatGPT revolution we've seen in the last 2 or 3 years. And so I think it's kind of like a microcosm of what we are to expect to come in other fields and other industries where AlphaFold, when it first came about, was thought to take the jobs of structural biologists. They were thinking, what on earth are we going to do? This thing can do everything for us, but they're not out of a job.

We're here at ISCB and there's lots of structural biologists, molecular biologists still working away. So I think it shows that there's going to be a lot for us to still do when the AI revolution comes to other areas.

Well, looking forward to meeting John. I think it will be an extraordinary discussion. Very excited. I think he's just over there.

We're now delighted to be joined by Dr. John Jumper from DeepMind. John, it's great to have you here.

Oh, it's wonderful to be here.

So I'd like to start off with this meeting at Liverpool. What drew you to ISMB/ECCB?

I mean this is a huge conference. What's amazing to me is how many people here are really great scientists that I know. I mean, the other is, of course, Dame Janet Thornton, who comes to me right after the Nobel and says, you have to come to ISMB, and I'm not going to tell her no. But, it's a wonderful meeting of the computational community, and it's really exciting to see. It's honestly also local for me. I came up from London. So it's just a wonderful meeting and a beautiful place.

And I imagine there's plenty of scientists here who have used your tool AlphaFold. So can you tell us a bit about how AlphaFold works?

So AlphaFold is this AI tool that predicts the results of a very specific experiment. This is determining the structure of a protein. And this is an experiment that lab biologists do commonly, about 11,000, 12,000 new structures are determined every year, but each one represents about a year of work for a scientist to determine, it's really an enormous amount of work. And these data have been collected kind of over the last about 50 years into the Protein Data Bank, this repository of these experimental results.

And we were able to develop new AI methods that work tremendously better at predicting the results of this experiment, predicting the structure of a protein for its sequence. This is something that biologists use all the time in order to plan their next experiments, develop hypotheses about disease, try and develop drugs that target a specific protein. And so the system that we made available has been used really worldwide by huge numbers of scientists.

And it's so heartening. I can't tell you how many times someone will come up to me, even at this conference, and say, we use AlphaFold every day and just so excited about it. And it's so important also because the work of science is so difficult, there's so much work that goes into experimental science, and it's really wonderful to be able to build a tool that help scientists do it faster.

Now, I think, I think I just, a carry on from it. I think it, it has revolutionized a lot of researchers' lives. I mean, I use it in my lab every day. But I guess what I wanted to ask really, have there been any examples that you've paid attention to, where they've used AlphaFold in a really kind of in a, in a really interesting way that really caught your attention?

I think there have been so many. I think one that's been really interesting for me is we kind of knew AlphaFold. You know, the story I just told, right. AlphaFold provides a prediction of a certain type of experiment. But what's interesting is people have started to find ways to do stuff that we didn't know it would do. One that I found really compelling is that people will use AlphaFold. They'll say, well, I don't know which proteins are coming together. What should I predict? Why don't I predict everything? Why don't I run AlphaFold 2,000 times and see what comes out?

And there was actually a pretty high profile result recently, how egg and sperm come together and the proteins involved in that interaction complex. And people ran 2,000 AlphaFold predictions, every protein on the surface of sperm against these known proteins on the egg. And they found one particular protein that AlphaFold could say this was the structure. And then they confirmed it experimentally. In fact, it was discovered independently by two groups computationally, and one showed experimentally, and it was a high enough profile result.

I remember I read about it in the New York Times and it, it mentioned AlphaFold and we, you know, had nothing to do with this other than building the tool. And here are people getting articles in the New York Times that isn't about AI. It's about this new discovery in biology using our tools. And that just absolutely kind of warms my heart.

So I just wanted to come on to a question about the Protein Data Bank. You've mentioned a few times how instrumental it's been. And as a, you know, part of our community with, we've got at it, you know, and gatekeeping it rather jealously, I guess. And I wanted to kind of ask how important was the Protein Data Bank and that repository of experimental data, how important was that for the development and the breakthrough that came with AlphaFold?

The PDB was so important, right? The AI is, of course, a tool that really brings together two things. You write a computer program, or kind of, or a mathematical formula, but really becomes a computer program. And this gives you kind of the skeleton of learning. And then you bring data and that actually fills it out into something that works. And you should think of machine learning or AI systems as code plus data yield something hopefully useful.

And so it's absolutely essential to have these data. And it's really the foresight of the experimental community that they collected these data, that you can't publish a journal article without getting the PDB to accept your structure, that there's very high quality kind of gatekeeping and curation that ensures that the data are available, that are reliable.

And of course, you know, it's very funny if you talk to computational biologists, if you talk to people at this conference, they'll say the PDB is the, you know, greatest, cleanest source of data we have, you know, what are we going to do after the PDB. And you talk to experimentalists and they say, oh, but you know, this is wrong and that's wrong. And that ligand was built.

Yeah, I know.

But based on hope, it's very funny to see these two. But they're both true. But I think that it's really a testament to the foresight of the experimental community, that they came together in a way that really we haven't in a lot of areas, maybe protein sequences and UniProt and others are another great example.

I think going forward, though, our data has to also keep track with how our science is changing. I think some of the most exciting, certainly in structural biology, are the questions of in situ measurement, how are we going to collect, say, tomograms, how are we going to collect, for, you know, people who don't know, these are flash frozen slices of whole cells. So you get to see the proteins in their native environment. And I think one really interesting question is what is going to be the equivalent of the PDB for slices of cells, possibly for different forms of microscopy?

I think these are some really interesting questions that, if we get them right, will yield AI models that make our lives better. And if we have to pull it from a thousand sources that are all somewhat different, it maybe we won't. And I think there's a real opportunity for data curation to make a difference in the AI systems we have in the future.

Can I ask you a couple of questions.

Certainly. That's what this is for.

AlphaFold, looking more towards in your career. You as a, as a scientist, we've done a lot of interviews over a long period of time with the physics community. And, and it's always bemoaning to the industry, it's always bemoaning to me, they're always saying that they really struggled to recruit the best scientists because the best scientists, you know, they wanted to work in, in kind of public science. Do you think your award is going to help with that in, in a way, of recruiting scientists into industry?

I think one thing that is very nice about the award is it's an acknowledgment that top tier science can come out of industry, and science that makes a really direct impact on the work of scientists. I think it's also the case, especially when you look in a field like AI, computationally, you know, computational power is important. It's the power to do the experiments, right. It's the, you know, GPUs are our reagents in a certain sense. And so I think industrial labs have had this big role to play in organizing the compute and organizing the teams.

I think one thing that is really a great opportunity in industrial research is that we can put together a team. AlphaFold was a team of about 15, right? We can put that group of people that are succeeding or failing together, that don't have the first author problem, that aren't trying to parcel you. You know, you do parcel out internally and you say, this person made this, this person made that. But as a group, you succeed or fail. And that's support specialization, that supports work. I think people learn a lot and they learn from really great collaborators. And I think it's a wonderful example of what this model can enable.

And of course, you know, it's always dependent, it's always part of your, part of a company. There's a lot that comes with that. Just like being part of a, in academia, there's a lot that comes with it. But I think it's a great opportunity. I think it really emphasizes also that great science happens everywhere, that, you know, leaving academia doesn't mean you're leaving the intellectual question. The research questions, and I guess maybe not even academia, but really the university system. I still consider myself a member of academia.

Yeah. You mentioned about data being incredibly important and then computing architectures being another kind of part of the equation when it comes to AI. And I guess everyone kind of had access to the Protein Data Bank. No. And we, we use still classical computers. It's not like we're using quantum computers. So what was it about your approach, your research, that essentially amplified that for you to get that value out of AlphaFold?

That's a great question. And I, I like to say, and I've said to audiences before, well, there's really three ingredients of AI. There's data, of course, there's compute and there's research. So there's the question of what is this computer program? And with the PDB, as you said, everyone had access to the same PDB. In fact, even more than that, over time we built AlphaFold 1, which was finished in about 2018. And we built AlphaFold 2 that was finished in about 2020. And we did not change the data between those two. We actually kept the exact same training data. And yet we learned tremendously more.

In fact, there was a really nice study by the AlQuraishi lab who retrained an AlphaFold 2 model on 1% of the available data. And they found it was as accurate or more accurate than AlphaFold 1. So you can say very cleanly that the research we did was about worth about 100 times the data, which is a lot of experimental structures. And so I think the right way to think about research is it lets you get more information out of each of these precious experimental points.

Now, within that, we had very new approaches and how we came to the problem, how, what kind of, what systems we built. We also developed this kind of team culture of incredibly high velocity, that we built on each other's ideas in a week, that someone would have a new idea and someone would build on that and have another and another and another. So I think we had this incredibly fast cycle compared to, you know, the de facto kind of distributed cycle of academia, in which a paper comes out. Other people read it maybe 3 or 6 months from now. It's tested in someone else's system, whereas for us it was next week.

So I think, and I can tell you a lot of stories of specific architectural innovations, but I think it's also really, I think a meme that sometimes takes hold is AI is this thing you take off the shelf and you throw it data, and you just have to have enough computers and you have the AI and you do it. But this is not true. That research is this enormous part, and especially in an area like protein ML, where I think it hadn't been taken seriously. Of course, proteins had been taken incredibly seriously. Machine learning had been taken incredibly seriously as academic disciplines. The intersection of protein and ML, there had been some work, but I think the degree of work wasn't as much.

We took very seriously, what is the science of this? How do we come up with new ideas? How do we test our ideas? How are we using the scientific method as we develop the architectural changes themselves. And we were able to do this, I think, incredibly quickly in all these ways. I think there's many explanations, but it is interesting. I think it is something of also the truth of modern AI research is that there really is, you know, everything out to these large language models. There's a tremendous amount of research and breakthroughs, and it's not always the top level ideas. There's all these little ideas in between that add up to a transformative system.

Yeah. And I think you've gone from this idea of having kind of lone geniuses, which are coming up with crazy theories in patent offices, to now collaborative teams.

Right.

Which is the future of science because of this interdisciplinary need.

I mean, I think, I think there's room for both, and I think there's, I think it's not only teams. I think teams work especially well when you have a clear objective that you're going after, you can get everyone to kind of agree. But I think also, you know, there are many ideas. I think one of the things that does change it is you need to test your ideas against what is currently the state of the art. And so I think that's one of the, like, you need to figure out.

We didn't have, we were maybe 30 points better on this GDT scale between AlphaFold 1 and AlphaFold 2. The single idea that gave the largest gain when we added it was worth two and a half. Most were one or less. And so you can think of us as kind of climbing, crawling our way forward and building. And I don't think it's incremental or engineering. Some of these ideas are brilliant, but we have to acknowledge that even brilliant ideas are worth a bit, and you still need many of them.

And I think we love to tell, you know, the story of lone ideas. Being, there was failure, then there was lone idea, then there was success. But even if you look at the history of anything, that you'll see kind of building on each other's ideas, that a much wider group of scientists than the lone genius brought us to where we are today.

One of the things you talked about in your, in some of the publicity you've done since winning the award is being, being a public intellectual. Are you, talk about being a public intellectual, which is interesting in its own right, but you also talk about that has a demand to do the right thing. So you tell me a little bit more about that.

So the Nobel Prize is wonderful. As I tell young students, I would recommend it, but, but it's, and it's wonderful as a symbol of science, as wonderful as a symbol of the promise of computational biology, that this, that we can do work on computers that matters for experimental scientists. And it comes with all these wonderful requests, you know, be it all the way out to, you know, talking to young scientists, talking to august bodies of science, talking to governments about this. And, you know, there's always this kind of invitation to speak as a grand intellectual.

Now, of course, you receive the Nobel Prize for a bit of science. And the, you know, the, you know, how do you understand computer architecture and machine learning of proteins? Not, of course, for government policy or anything else. And so I think there's always to me this responsibility that you want to, that it is a platform that you should use to do good when you can do good, that when there's an opportunity to come to a wonderful event, you should generally try and do it.

And at the same time, you should remember that you too are human. That this is a prize that recognizes an accomplishment, that I have a lot to say in the design of structure prediction algorithms. I have maybe something to contribute to the organization of science, but it needs to be done with a lot more humility than I talk about details of protein structure prediction architecture. And so I think it's really balancing the opportunity with the humility of what do you really know? What can you really contribute to these big, important questions?

And as you said, the most important job's being the father, right?

Oh, very much. And that's the other is, yeah, I have three young kids, you know, six, nine and 11. And I want to, I want to be incredibly sure that they are not shortchanged of having an active, engaged father because I'm busy doing important things. And I think getting all that balance right in, you know, being in this research area that's still incredibly, you know, quickly developing, still incredibly important. And you want to contribute, you want to go, you know, there's a good portion of me, my wife says if it was just up to me, I would just hide away and do science. And it's pretty true. But, but I think all of those and just trying to find the right balance, that, you know, the Nobel's a wonderful acknowledgment that my career has done some good for the world, and I would like the remainder of my career to do some good for the world, too.

So just moving on to the remainder of your career, what, what is, obviously the, the sort of long term view, but what is next? What's the next big question that you're hoping to answer now? AlphaFold has been so substantive in that leap forward.

I think so. AlphaFold and technologies like it, and technologies, and other technologies will continue to grow and develop. I'm actually very interested in, you know, can we find the next question? Can I kind of find the next thing that will happen that doesn't look as plausible now? And actually, honestly, about a year ago started to get into this question of we're developing language models, chatbots, things that can clearly read and write science, they clearly kind of understand.

I remember one of my first kind of surprising experiences. I just gave it a paper in the context and I said, you know, list me all the actual experiments performed in this paper. And it did a fine job. And I thought to myself, that's, that's non-trivial, right, to, to kind of pull a dense scientific paper. And, but the real question is, but if you ask them kind of a detailed question like what will happen in this experiment or something else, or, you know, something that you really know? Well, they tend not to do so well.

And I think I'm very interested in this question. What gets us to biological reasoning? To what extent can we make language models understand science in a deep way, engage with it as kind of as a scientist perhaps does, is, and I won't say too much about how we're trying to do it, because I haven't succeeded yet. I got to keep some secrets. But I'm very interested in this question of how are we going to develop reasoning?

Because there's so many problems that don't have a PDB, that in fact, even if you collected all the data in the literature, wouldn't be much. So we're really going to have to learn the grand trick of machine learning, is to take a, you know, a question you're interested in and take some data that looks mostly unrelated and make it just related enough, to make a protein that you want the structure of and find another protein in a completely different family and learn a bit from that that you can apply here. That's really what is amazing. How do we do this in scientific reasoning, I think, is one of these grand problems that, you know, people express much more doubt when we started playing with this a year ago than they do today, of course, due to the rapid development of these technologies. But there are enormous challenges ahead.

Yeah, of course. And I can imagine if you get a model good enough, then I guess it could highlight the gaps in our knowledge and places to go looking. I guess that would be the ultimate aim.

I mean, there are many, many things that you can do if you actually get good at understanding this. We will have gaps in our knowledge, will have some understanding also of what's implied by their experiments, ability to find really relevant kind of literature. There's much that we can do, but we need to, you know, sit down and not just dream, but find the, the local part of how do we get from where we are to, to this future when these tools are really effective and trustworthy?

So, so one of the things I found very interesting, again, in some of the, the interviews that you've done, when you talk about the speed of AI and the speed of, I wouldn't say evolution of AI because potentially not even evolution. And that brings me to two questions. One is, you know, where you, you know, where you see things going in the medium term, but also public trust. You know, there's, there's already a shortage of public trust. Most people don't understand what we're talking about here today. And it goes faster and faster and faster and more gets done. How, what do we need to do to get people to trust AI?

Really in terms of trust, it's been a very interesting, I think one of the interesting experiences for me is personally been AlphaFold, and I were, and we weren't sure in a lot of ways, you know, even when the, our CASP results came out, that was this public assessment that said that blinded we could predict structures. I still think there was a lot of doubt in the experimental community. Well, maybe those are easier. Or maybe, you know, it wasn't until about six months later when our paper came out and our code was available and predictions were available, that people really started, you know, the predictions were available on a website. So they said, I'll just check.

Yeah. Yeah, yeah, yeah. We checked.

Yeah, they checked. They said, or, and a lot of people had solved an experimental structure and hadn't yet deposited in PDB. And so they said just, you know, just for fun, I'll go see. And I remember someone on Twitter saying, how did they get a copy of my structure? Right. You know, did you send it by Gmail or something? Right. There was absolute disbelief within the community that suddenly these methods all at once were really relevant.

And I remember when we were putting this out and we were actually talking with EMBL-EBI and Sameer and others, like, how do we present this information? We were worried about? We weren't as worried that people wouldn't trust us. We kind of figured once they really had access to the predictions, it would take some time. I thought it would take more time than it did, but people would learn it. We also worried they would trust it too much. How would we get them to doubt it? Just the right amount.

And we'd put a lot of effort by this point into what we would call confidence measures, where the model predicts how far it is from the answer in, in various dimensions. And that, and I think one of the better, two of the better decisions we made. One of the better decisions was, you know, we would go, we took this kind of numeric confidence measure called pLDDT that we had developed. It was 0 to 100, and we decided to just define four cutoffs. Right, above 90, we would color the structure dark blue. You know, less than 50 would be red for danger. Right. And we had these kind of four things that were very clearly aligned with what people should do with this. Probably trust, trust it quite a bit. Probably trust it, probably don't trust it. Definitely don't trust it.

And the experimental community took to this and others really well. I felt like they really said, well, these are imperfect. We know they have errors, right? We know they have errors in part because people have seen them. But also we know that the errors tend to be well correlated with when the model says it doesn't trust it. And then we'll kind of do this, we'll look at it, we'll think about it, will design our next experiment. We'll test aspects of this prediction. And they very quickly kind of integrate it into the scientific practice, trusting it in about the right way.

And I think the reason this happens so quickly is the scientific community always deals with uncertainty. They never fully trust another experiment. They never fully trust their own experiments. They're always thinking about all the ways. I mean, you're an experimentalist and, you know.

Yeah, absolutely. We, we, we don't trust our own hands sometimes.

Exactly. And so they're used to this kind of strategic doubt, doubting in just the way that still lets them make progress, just enough skepticism that they aren't misled. And they can, but yet enough trust that they can still make progress. And that integrated really well into how scientists work. And I think the larger question of AI is people developing over time this understanding of what these outputs of a large language model mean, what they are useful for and what they're not.

I personally see some parallels. I grew up, I was born in 1985, so I grew up kind of, my transformative experiences were in the 90s when the internet was new, and there was AI early on in this, you know, how can you trust the internet? Anyone can put something on the internet. And of course, that's true. And of course, there's a lot of really dumb things said on the internet. But also we learned how to use and navigate this tool. How to think. I remember being warned about Wikipedia. Anyone can edit Wikipedia, right? Was what they would tell every school student in the 90s. And we learned to kind of navigate that and still obtain a lot of use out of it.

And I think we'll see something similar that, you know, AI became a really relevant force very rapidly. And society, I think, will adjust to the strengths and weaknesses. And it won't be, you know, it won't always be easy. But I think that this will come, and I think hopefully we'll be able to develop more measures, certainly in the scientific space, confidence measures. And I think this notion that we had in AlphaFold, that the way you get to know a model's confidence is just to ask it to predict how wrong it will be, was itself kind of a fun fact, right? That's a weird, you can get a model to tell you it's wrong without telling you what the right answer is, and I can, I can kind of give a human analogy in terms of you often go out of a, come out of a test knowing if you did well or poorly without knowing what all the answers are.

But it is very interesting. I think we'll see it develop over time, but also the kind of world and scientific systems will adjust. And I certainly have seen some bad uses of AlphaFold in science, but overwhelmingly, I think people using it really well and developing the right techniques and sharing in the community. Right. It is really the community that learns how to use these tools to do their work better.

And I just want to follow up with regard to the scientific process itself, where traditionally it's all been about explanations and understanding mechanistically what's happening and then kind of getting to the solution. Do you think it's okay if we solve, we find the solution to the problem without even understanding it? I mean, if the goal is just solving the problem and we solve it and we don't really know what happens in between, particularly in the context of AI being a black box, which is difficult to understand. Does that really matter? And how does that kind of change science?

I think it's an interesting question, but I have kind of two perspectives on it. And I would even say that science isn't about explanation. It's about testing, that we have a hypothesis, we deduce a conclusion from it, and we see, especially surprising conclusion, we see if that surprising conclusion itself is true. So in that sense that transfers quite cleanly over to having, you know, something like black box, maybe not explanations, but systems that make predictions.

And in fact, this is not a new thing. We've been doing computer simulations for a long time, and you can explain how you came to this computer simulation, but then you run a million, a billion steps of this computer simulation. You get predictions and you test those, but you don't in practice have an explanation of, in a molecular dynamics simulation, you know, how did we get to this? And of course, that hasn't been a huge stumbling block for science. We have theories of how you build simulators. We have the results of running simulations, and then we integrate them into our larger practice. And ultimately we test the consequences.

So right, when we do this alpha, we don't want protein structures as the final goal. We want them because they help us understand biology, that they help us make new hypotheses. And then we test those. When we make drugs with computational inputs, we test the drugs, not the computation. Right. I think that's not a fundamental challenge.

I think also there is a science here, but the science isn't, is, you can't cut it quite in the middle. You can't say aliens handed you a black box. And now science, the science is we have an understanding of how machine learning kind of develops understanding. We're developing kind of localized science in terms of how do we make protein structure prediction systems, what can we change, what matters. And then this overall loop of data plus machine learning yields a predictor that you then use that has a clean explanation and a science.

But it's a little bit, I think we're, I think it's in part because people kind of don't see the first part as much, or it's done by fewer people. And so, but they see the artifact that, that they have the challenge. But I think ultimately we'll just continue doing, we'll continue making predictions. We'll continue deciding whether those predictions are general, a really generally reliable, the same way we've done with theories and with computation forever. And I think it will, it will not show up as a major kind of challenge to the paradigm. Maybe later if we have systems that are hypothesizing and testing them. But right now I think it fits nicely within the scientific paradigm.

I just want to say thanks ever so much indeed for taking time out to your schedule to talk to us. We've really appreciated. It's been such a fascinating discussion for us. So thank you very much.

Thank you very much.

So guys, John Jumper, what an interview. And what a gracious guy.

Oh, phenomenal guest, right.

I didn't quite ask him if he had his Nobel medal on him time. I kind of wish I did though. I wanted to see it.

Yeah. I thought the discussion was with John was really interesting because I think all we hear at the moment is that AI is about brute force. Whoever has the most computational power is, you know, going to make it. And I think him emphasizing the importance of research, and yes, data and compute are important. And of course, Google has a lot of compute. But the research really was the key to the success of AlphaFold.

Yeah, I think, I think the, the idea of the scientific method still reigns supreme in research. The idea that you can generate this black box that might produce a result, but you still have to validate that result to really find out how truthful it is. And, and he did that. And he talked about that with his own AlphaFold, sort of incremental, to, to make it more accurate and more powerful. And the same is true for the protein structures that come out of those predictions. You're still left with the need to validate. And he clearly confirmed that when he was here.

So that was nice. And he had a lot of compliments to pay about the work that you and your colleagues have done in that whole areas.

And yeah, I think, I think he really clearly values the Protein Data Bank, and, and the fact that he commented that it was possibly one of the cleanest data sets you could train a large language model on, I think speaks to the need to, to, to produce and develop other, equally good data sets for questioning other areas of biology. And he hinted at that with the next iteration of thinking about ways to get AI to think in biological terms, where is it going to get the data from in order to be accurate?

And I think we have to think about now what is the next version of the Protein Data Bank, but for other biological questions. But even beyond biological questions, material science, physics, chemistry is, and actually sales.

Yeah. Business, financial documents.

But how do we build that into our systems, our processes, so that we are collecting, one, the right data. We're collecting the clean data as well. And how are we doing that in a way that's easy for our systems, our organizations, our groups.

Fascinating discussion. Great guest to have on, wasn't it? I really, really loved it. And, and what a show, right.

What a show. I mean, it's one of the best. Say, I think it's going to live in my head for a long time.

Anyway. Well, thank you guys for joining us today. Luke. Stephen.

Thanks for having me.

What an amazing show we had today. Thank you for joining us. Please like and subscribe if you'd like to see more content like this.