Should we be scared of AI viruses?

AI-designed viruses and lab safety

0:00This BBC podcast is supported by ads outside the UK.

0:30That's healthy optimism. That's Optum. Visit Optum.com to learn more. Today's episode is brought to you by ChatGPT for Business. As a listener of this podcast, you're looking for ways to help teams move faster, make sharper decisions, and turn scattered context into work they can use. ChatGPT for Business can help. ChatGPT for Business gives teams a shared workspace with admin controls, permissions, and access to work and codecs in ChatGPT. This means your business can move from question to answer, and code to rollout quicker.

1:03Join over 10 million business and enterprise users worldwide already using ChatGPT for work. Download the ChatGPT desktop app or contact sales to learn more.

1:15Welcome to Inside Science on the BBC World Service. I'm Ronan Pease, and today we'll be thinking about NASA's two remarkable Viking space probes which touched down on Mars 50 years ago, breaking frontiers in space science. We'll reconsider the impact of their pioneering images of its arid landscape, and take a look at how today's ambitious successors have built on their success, and the sweet, crisp sound of desiccated leaves. No, we're not returning to the drought again.

1:48Rather, these crumbling leaves hold the secret behind a new fridge-free vaccine that listener Peter from Bristol wrote in to ask about. It's familiarly known as the resurrection plant. Although it appears to be clearly dead, it's not really dead. It's in a state of suspended animation. If you put this thing in water, it will expand into a bright green living plant. And I'll keep you in a state of suspended wonder how on earth that relates to vaccines. And joining us with a different perspective of what's going on in science and technology,

2:23Gareth Mitchell. Thanks for joining us again. Lovely to be here, Roland. And I'll be looking at what happens when artificial intelligence sets itself free. Well, artificial intelligence is in fact where we're going to start. But actually, moving over to biology, specifically the area they called synthetic biology, the extreme possibilities of which hit the headlines last week when researchers at Stanford University said they'd created new kinds of viruses that had been designed by AI,

2:53which has the potential to sound rather scary. Think COVID or thoroughly exciting. Think biotech, depending on which way you view your glass of genomes. I'll be looking at the ethics and the risks of this kind of work in a moment and whether new regulations are called for. But first, the science of what they did. What does it mean to design a new virus? And what does AI bring to the process? Microbiologist Simon Clark stepped forward to explain, reassuring me first that this was done with viruses that cannot infect me or you.

3:28They only infect bacteria. And that this is a discussion we were bound to have one day. This was going to happen anyway. And sooner or later, somebody would have tried it. So it's not the case of they've let a genie out of the bottle that should never have been allowed to escape. They've told us that it's there because I suspect otherwise people might have been tempted to dismiss it as sci-fi and far-fetched. And this shows that it's not. It is a thing you can do. They chose to do, as I understand it,

3:58a very small and easy virus to start with. Yeah, that's right. They've actually created 16 new viruses. And these are a virus that infects bacteria. They do not infect humans or animals or plants or anything like that. So they have no effect on human health or agriculture. But they are useful in a lab setting. And that's what they've done. They've used artificial intelligence to produce a bunch of these new viruses. And when you say they've used artificial intelligence, presumably it's trained on known viruses,

4:35maybe even known versions of this particular virus that they're studying. And at that point, they ask it to sort of riff on the idea, sort of riff on the theme, a bit like a composer might do. Yeah, riff is a good way of putting it. They've asked it to look at millions, probably, of genome sequences for different other types of virus that infect bacteria, different bacteria. And it's come up with thousands of possibilities. And from that, they've filtered the possibilities.

5:05And they've come up with about 300 of these. Only 16 of those turned out to actually be viable when they were put into bacteria. I should say, of the 16, only four were able to outcompete the original parental virus. So while it makes new versions, they're not always better. It's very interesting. I mean, in the sense that they've created new viruses sounds very scary. In the detail, in fact, there's a lot more nuance about it that these aren't completely different from anything seen before.

5:42They're not necessarily better. But what are your thoughts on why people might want to use this method? The researchers here have hung it on the hook of overcoming resistance and conquering antibiotic resistance, because some people suggest that you can use these bacterial viruses to kill antibiotic-resistant infections. The problem with that is that the bacteria become resistant to the viruses as well. So they said they've generated the new viruses in order to overcome that.

6:15I'm a little bit sceptical there. But that's the good use of it. The bad use of it is that this demonstrates that if you've got enough copies of, say, genomes of, say, flu, you could perhaps generate new, more transmissible versions of influenza, that is possible now. With the right model, it is possible, and with enough genomes. Looking at the paper, this would, in my reading, fall under this rubric of gain-of-function,

6:49where you change a genome so much that it's doing something different from what it did before, which is being banned in the United States at the moment. Would you agree that this is some kind of gain-of-function work? Oh, yes, definitely. They have made fitter versions of this virus, so you've gained function. Yes. The ban in the United States, as I understand it, is that this is a ban on dangerous gain-of-function research, not on all gain-of-function research.

7:19And because these viruses absolutely are no threat to human health, you can't really say that it is dangerous gain-of-function research. And just to be clear, you say it's not a threat to human health because the cell biology of human tissue is so different from the cell biology of bacteria. Yeah, they do not. They will not infect our cells. They cannot use the replicated machinery within our cells to generate more copies of themselves, which is all a virus really exists to do.

7:50So they are no risk to human health at all. Dr Simon Clark of Reading University.

Ethics of synthetic biology research

7:55Well, I don't know if that was reassuring or just raised the alarm levels further. So I'm joined now by Joy Zhang, who's a professor of sociology at the University of Kent, specialising in the ethics of biological research. And she's also in-house ethicist for an even more ambitious but longer-term project called the Synthetic Human Genome. So, Joy, first reactions to these AI-designed viruses? I think this is actually an exciting milestone in a very long and collective process of the scientific community

8:28trying to work out how to understand and write genomics. So what, you're saying that how you can deal with a virus's genome, how you can manipulate that will have bigger ramifications down the road? I think this is more of a proof of concept of what we can do with AI, with a very simple set of genomes and looking at bacteria. But nevertheless, it's a great achievement. I mean, as I was talking with Simon Clark,

8:59I think people are likely to be quite alarmed by the idea of manipulating a virus in this way, and that the creativity of AI makes that even more worrying? I think we should be very attentive to how this field is developed. I think a healthy amount of worry is warranted, but I wouldn't say we need to be panicked or be doomistic about this, partly because the research is on phage, and the reason I think scientists love to work with bacteria phage

9:32is because it's very simple and it has huge therapeutic potentials. One of the things that Simon mentioned was that this sort of opens a prospect of engineering flu in a different way. I did go and check. The flu virus has a genome that's only a little bit bigger than these phage viruses. So if they can, you know, with AI, test out a whole load of potential viruses and then cook them up in the lab with one of these bacterial viruses,

10:03then how far off would it be to do something like that with a virus which would be much more concerning like flu? I think you highlight a very important point. That is, people are worried about what about misuse or abuse of this technology. And so I think the right question to ask is not so much about how we're going to contain it, how we're going to restrict access to a technology. As Simon said at the beginning, someone's going to do it. And also, we actually, as a community, we want to explore different scientific possibilities

10:38so as to prepare ourselves for the next pandemics or other global challenges. And more importantly, governance, especially good governance, is not just about restricting conduct. Governance is about conduct of conduct. That is, how do you steer scientific inquiries? How do you steer innovation into the right direction? It's not just about how we block the bad things. It's about what kind of incentive now we should think of so that scientists are more inspired

11:08and incentivized to use AI to do the kind of research that can benefit public health, public well-being, and possibly ecological health as well. I mean, what are the arguments you hear for doing this? You know, if you're saying that, well, we have to think about the downsides, there presumably have to be upsides if, as Simon said, this was something that was bound to happen. I think for this specific research that's published in science, I think the intention has huge therapeutic potential

11:38and is about bacteria. I think it's quite exciting. And of course, as a social scientist, I start to think about the implications of further down the road, if we're moving beyond this specific case to the wider spectrum, what may happen? And I think this is a question that all of us should think about, not just social scientists, regulators, scientists as well. Joy Zhang, thank you very much for joining Inside Science. Thank you. So in the studio with me is science journalist Gareth Mitchell.

12:09Gareth, what was your reaction when you heard the story about an AI virus? I mean, cautiously excited. I mean, it really is a step forward. But of course, there are dangers. And let's hope that the ethicists and the regulators and so on are going to keep us safe on this one. But if you look at what has actually happened here, you know, this is proof of principle that AI can basically can go off script and it can create completely novel genomes. And not just genomes, but viable ones, you know, that they've shown in this research, they can create them in a lab, put them in a Petri dish,

12:44and they infiltrate an E. coli bacteria, you know. And that in itself, I think, is a stunning achievement. Well, obviously, we're going to be talking about this more in the future.

AI agents breaking their sandboxes

12:54I want to turn, though, to something else, which I think you've been looking at for us, because the big concern about these viruses is they marked escape the lab. And there's been a lot of talk about AI agents, I think is the word, that have been escaping their sandboxes. You better talk me through this. Oh, wow. Yeah, they break free. Because, I mean, to talk about this so-called agentic AI or AI agents, just so, you know, most of us, I think, are familiar with large language models like, you know, Google Gemini or ChatGPT.

13:25They'll give you an answer to a question, but they don't actually do anything. Whereas an AI agent is something that any one of us could use on our computer, for instance, to automatically go through our accounts, maybe. But the stories which I'm worrying about, the ones where they've gone and hacked sites and... This is it. Yeah. So we've had a few cases the last few months. I opened AIs. AI hacked into this AI repository called Hugging Face. One of Meta's AI models managed to kind of get out of its so-called sandbox.

13:55So just to say these agentic AIs in a research environment are grown in a place where you think, well, they can't interact with the outside Internet. But in this case with Meta AI, it did. And then there's one that's been all over the news this week. You've probably heard about it. A guy in Melbourne just got agentic AI to book an appointment for a Pilates class and just said, you know, can you just get me a bit further up in the queue? And the agent says, yeah, I'll do that for you. And the way it did it was actually to hack into the gymnasium's whole booking system and just kind of hack this person's place into the queue.

14:29This is AI breaking loose. And what's the solution? It's a worry, isn't it? And so what this is down to is that where human values and machine values clearly misalign, you know, and I'm sure the guy in Melbourne didn't want the AI to go quite that far. So what this is leading to is a whole discussion about how you regulate. You have this kind of reinforced learning with these AI systems so that you have a human in the loop idea. So these agents don't just go off and do loads of other stuff.

14:59Gareth, you're making me quite dizzy. Me too. I mean, we'll see where that goes.

Revisiting NASA's Viking Mars missions

15:04What I do want to do, though, now is to turn back 50 years ago this summer when NASA took a giant leap in exploring the other planets of the solar system. Good evening. This is Mars. Incredible pictures sent back from Viking showing a red, rock-strewn landscape under a pink sky. The detail is absolutely amazing. And these pictures would have seemed science fiction not so very many years ago. Viking is now standing in the plane of Christy. It's working. It's sending back information at this moment.

15:36Ah, classic sky at night. And I sort of remember it, those landings of NASA's two Viking missions on the surface of Mars. What we all wondered was whether there'd be Martian bunnies hopping around between the rocks or at least some kind of fungal growth. There wasn't. It rather just looked pretty desolate. And most of us moved on. Even NASA ignored Mars for a couple of decades. But this was, in truth, a remarkably bold mission that gave us the first science from this alien world.

16:07And it laid the groundwork for the two remarkable rovers that are now wandering over Mars' surface, sending back amazing detailed images of landscapes that were once shaped by water billions of years ago. Maybe even hints of possible early life. Those are run by NASA and Claire Cousins, who joins us now, is part of a team that's hoping to put a European rover on the surface. Before, as they say, the decade is out. Thank you so much for joining us, Claire. I suspect you don't remember the disappointment from those first Mars pictures.

16:41But are the Viking landers actually treasured amongst real scientists like you? Yeah, that's right. I, in fact, wasn't even born when the Viking landers touched down on Mars. But yeah, as you say, they very much paved the way eventually for future Mars exploration that was then going to happen. Despite that hiatus and exploration that you mentioned, you know, once Mars exploration did pick up again, it was very much building on the lessons learned from those landers. I mean, were those sort of technical lessons about how to operate on Mars?

17:14Or were there scientific lessons about, you know, the composition of Martian rocks or, you know, that kind of stuff? It was both, really. Every subsequent mission that gets sent to Mars, you know, uses heritage and experience from the technology sent from previous missions. But scientifically, there were huge lessons learned from the Viking landers, particularly to do with, you know, how to look for life, right? You know, so, yeah, no bunnies, no fungal growths on Mars, but also no signs of bacterial, you know, microbial life forms either. It seemed to be completely devoid of any kind of biological process.

17:46And this was very disappointing at the time for the mission and that was hoping to find evidence of present day life on Mars. But since then, you know, we now know actually what the Viking landers were trying to achieve was really guided by the science that was not sort of up to date at the time. You know, it was looking for kinds of microorganisms, you know, that utilize sunlight, for example, that would use respiration to get energy. Whereas actually now we're so much more aware of the wide variety of microbial life that occupies our own planet. You know, life that can live in extreme environments, life that can utilize energy from minerals and rocks, for example.

18:23So now we have a much better idea about what to actually look for as well as where to look for it, too. And also looking mostly for fossil life of some sort and, you know, the realization that there are these water shaped landscapes, which, for example, Perseverance is exploring at the moment, one of the NASA rovers. Yeah, exactly. So, you know, far from this, you know, kind of desolate wasteland that the Viking images revealed to us 50 years ago. Instead, now we actually understand that early Mars was really not entirely different to what early Earth would have been like, you know, three or four billion years ago.

18:59And we can then look for evidence of past life that would have perhaps been living on Mars early in its history rather than any present day life, which I think everyone's in good agreement that, you know, that's not there on Mars today. So tell me what a European lander would bring to all of this and what's your part in all of this? So the European lander, it's a European Space Agency-led mission. It's called Rosalind Franklin. That's the name of the rover. And it's going to be landing in an extremely old part of Mars. So the terrains it's going to be exploring are about four billion years old.

19:31So we're kind of stepping back in time to explore this early history of Mars. And one of the really unique things about this mission is equipped with a drill, which will actually be able to drill a couple of meters down into the Martian subsurface. For Mars, this is a huge first. It means that we can actually access geological material that's been hidden underneath the surface and protected from surface radiation, which constantly bathes the Martian surface. And that destroys things like organic molecules. What we can detect are things like residues left behind from those life forms,

20:03which have been captured in those rocks, kind of squeezed between minerals and then preserved for this long period of time. And we can also look for geochemical traces left behind in the rocks as well, which life may have influenced. So it's going to be a bit of a mixed bag of evidence if indeed we do find anything. I think you were supposed to be on Mars already, but the Russian invasion of Ukraine put that all back. It was going to be a Russian launch system. How jealous are you of your counterparts in America who are getting these pictures back from curiosity and from perseverance?

20:38So, I mean, it's really wonderful to see that these these rovers are so successfully exploring the Martian surface. So, yes, on one hand, it is frustrating that Europe has had several setbacks over over the years to getting our own rover landed on the surface of Mars. But in the meantime, it's still just as a Mars scientist, so exciting to see what is continually coming out of the missions that have landed on Mars. And of course, those that are in orbit around Mars as well sort of supporting those missions. So really, this just constantly increases our understanding and allows us to be better prepared for when we do finally get there.

21:12And which will be when? So we're launching in late 28th and then landing in 2030. So still a few years to go, but this gives us a lot of good time to improve our preparations for the mission. Well, thank you very much for telling us about it, Claire. And fingers crossed, I should add that Claire Cousins is a reader in the School of Earth and Environmental Sciences at the University of St. Andrews. Thank you so much. You're listening to Inside Science from the BBC World Service. Tell us what science you think we should be investigating.

21:43Our email address is insidescience at bbc.co.uk.

21:51Healthcare can feel complicated. That's why Optum uses technology to connect the people and processes that make healthcare easier, more affordable and more effective. We're making it clearer for you to know exactly what your benefits cover. And to help you better manage your health, we're coordinating care between your doctors and your technology. We believe better, simpler healthcare is always possible. That's healthy optimism. That's Optum. Visit Optum.com to learn more. Before you jump back in, you know that feeling when feedback, notes, and data are coming in from every direction?

22:28ChatGPT Work can make sense of it all, turning scattered inputs into ready-to-review work. With your permission, ChatGPT Work can gather context from apps and files, summarize incoming messages, rebuild decks, and help finalize high-quality documents for your team or clients. Stay in control as you move from a goal to polished work faster with ChatGPT Work. Download the ChatGPT desktop app today and take on your most ambitious work.

Resurrection plants and fridge-free vaccines

22:58The other day, this question came into our inbox at insidescience at bbc.co.uk from Peter in Bristol about some other science news. Dear Inside Science, the recent news on the radio heralded a breakthrough in the preservation of vaccines without refrigeration. Is this genuinely new? I recall a similar claim of 10, perhaps 20 years ago, using concentrated solutions of various sugars.

23:29Snap, Peter. Because my recollections went down precisely that rabbit hole. And it wasn't 10 years ago, nor 20, but 31 years ago that I went as a fresh-faced BBC reporter to the Quadrant Labs in Cambridge to meet the resurrection bush that inspired that recently approved fridge-free vaccine. It's familiarly known as the resurrection plant. It comes from the deserts of the southern United States and spends most of its time in this dry, desiccated condition.

24:03In fact, it's so dry that Texas cowboys use them as kindling to light their campfires. The interesting thing about it is that, although it appears to be clearly dead, it's not really dead. It's in a state of suspended animation. Just add water and the thing comes back to life. Who you heard there was chemist Bruce Roser, who persuaded me right away that the magic ingredient in the resurrection plant, a sugar called trehalose, would encapsulate and protect fragile vaccine molecules just as it does the resurrection bush proteins.

24:38And he even gave me a taste of the trehalose.

24:43It just tastes like ordinary sugar, really. Not quite so sweet. Not quite as sweet as ordinary sucrose. We found that we can stabilise virtually every vaccine that we've looked at, including all the childhood vaccines. And these can now be stored outside of the refrigerator so that they can be delivered in the third world into local villages and administered without the requirement for refrigeration. And it's the same Bruce Roser who, with his newer company, Stable Pharma, has at last brought that lab dream of trehalose stabilised,

25:15no-fridge-needed vaccine to clinical approval. An idea preserved like a desert plant for decades. But, Ronan, do we know why it took so long? I know. I mean, I've always been waiting for this. It seems like such a good idea. The team here at Inside Science went and did some digging around. And what they heard was that, essentially, the pandemic changed everything. Suddenly, ideas that had been on the shelf. Think of the mRNA vaccines waiting their moment. They came back. And it looks like that's the story with trehalose as well.

25:47But, you know, I remember, I've been reporting on science for 25, 30 years. And, like, LEDs might seem a bit boring. Light-emitting diodes. But how they've come on in 25, 30 years. Yeah, that's interesting. So, the blue LED that came out in the 90s, that was the one real example I ever turned to, where we reported something, and a year or two later, it was being sold everywhere. And it's been absolutely transformative. In a sense, it all shows, A, how hard it is, actually, for us to make the right call. And sometimes, you know, like the trehalose sugar, years and years and years on, there it comes through.

26:21Resurrected. Yeah, absolutely.

Navigation, maps, and brain health

26:23Okay, I can give you time for just one story, Gareth, that you've picked up when you've been scanning the horizon this week. Right, okay. So, on the horizon, I've been looking at the conversation written by academics, not journalists. So, you get these whole other perspectives. So, this was an article that was really asking, like, if you're using navigation skills, you know, your kind of spatial reasoning and so on, working your way around, like, a map or a dense urban environment and so on, is that good for your brain?

26:53And then it cites a few different studies. For instance, in 2024, there was a study in the US that found that ambulance and taxi drivers were less likely to die from Alzheimer's than people with other kinds of jobs. I mean, it was quite a small effect, I have to say. But there was a bit of an effect there. Then, probably a lot of people remember this. Back in the year 2000, there was a study into London taxi drivers who, of course, famously memorised the map of London and what they call the knowledge. And this found that their posterior hippocampus in their brain, which is an area that can be,

27:27and is often attacked early on in the progression of Alzheimer's and is crucial for navigation, for instance. But in these taxi drivers, in brain scans, it showed that they had, like, an enhanced brain area there as well. So, something might be going on. Now, posing this question actually turns out to be a professor of civil and environmental engineering, not a neuroscientist. Right. You might end up, well, what's this person? Why is he interested in this? Well... And this is because he works a lot on geographic information systems. So, he spends a lot of time looking at a screen, maybe mapping cases of disease against different geographical areas.

28:02And he was thinking, well, I'm using my navigational faculty a lot. Maybe this is quite good for me. And he's just throwing it out there as a research question. Do you need to be driving a taxi all day to get this benefit? Or could maybe a kind of geospatial civil engineering kind of person have some kind of benefit? But, Rodan, what I want to leave you with is a question that I had, which is maybe it tells us all that we should be less reliant on our apps for our mapping apps and so on, our sat nav and so on. And maybe we should just stop following that blue dot everywhere and try and navigate for ourselves.

28:32Of course, I'm not claiming in any way it's going to ward off neurodegenerative diseases, but it might just be good for our brain. Thanks very much, Gareth. Always good to have you navigating us through the corridors of science. And that's it this week for BBC Inside Science with me, Ronan Pease, here on the BBC World Service. Tom Whipple-Whoop will be back next week. I do hope you can join him. The U.S. has long been suspected of trying to sway elections in other countries, but is something more brazen going on these days?

29:04I'm Asma Khalid, and I'm one of the hosts of the Global Story podcast from the BBC. Brazil has accused the U.S. of meddling in its upcoming vote. The U.S. has dismissed the allegation. But how do these claims fit into the long legacy of American electoral interventions? Check out The Global Story on BBC.com or wherever you get your podcasts.