Flu vaccinations, and calls for AI regulation
As flu season starts to bite the Northern Hemisphere, we look into the efforts to develop the most effective vaccines. Also, an AI expert reviews the recent Safety Summit hosted at Bletchley Park, how chimps are demonstrating human battle tactics, why cockney accents are becoming less common amongst young people, and how one might go about bending a laser around the Moon...
In this episode
01:00 - Machine learning helps doctors stay ahead of the flu
Machine learning helps doctors stay ahead of the flu
Derek Smith, University of Cambridge & Jessica Stockdale, Simon Fraser University
In the northern hemisphere, flu season is once again upon us. Each year, as the virus circulates, millions are infected and hundreds of thousands of people die from the infection, especially the more vulnerable. Thankfully there is an international vaccine programme which uses samples collected from the opposite side of the world during their flu season to try to stay one step ahead of how the virus will evolve and change. Cambridge University’s Professor Derek Smith, an expert on the evolution of diseases like the flu, sits on the panel that designs those vaccines…
Derek - Flu is a respiratory virus. It kills, worldwide, about half a million people a year. It infects about 10% of the world's population each year.
Chris - Why do we have a flu season? What drives the fact that it always comes at a certain time of the year, give or take?
Derek - Well, we don't know. In places like the UK in the northern hemisphere where there's clear winters, it happens in the wintertime. But flu happens in Bangkok where it's 37 degrees year round, pretty much. In tropical regions like that, it usually happens during the rainy season. Now whether or not it's the amount of moisture in the air, whether it's the fact that during the rain and during the winter we're all in closed rooms together and we can transmit more easily, whether we don't see as much sunlight, it's really unclear and it's been an outstanding question for many, many years. Nobody knows yet.
Chris - And why do we keep catching it? Because if I catch measles, which I luckily haven't, I'd be immune for life. Why is that not the case with flu?
Derek - It's not the case with flu because the flu virus has adopted a lifestyle where it can change over time to evolve, to get mutations that can escape the immunity that we get. And typically we all get flu once every five or 10 years.
Chris - When it changes, what does it change about itself to mean that I can keep catching it?
Derek - It changes in a way that escapes the antibodies that we've built up against earlier strains in a very similar way to the antibodies that we all had against the early Covid strains, either from being infected or vaccinated. Those new variants as well, they've escaped that prior immunity, and flu does the same thing. There are critical amino acid substitutions on its surface in the area where the antibodies care about that can make it so those antibodies we have can't bind there anymore and neutralise the virus. So these new variants can reinfect us because our old antibodies don't work anymore
Chris - And that's why we need a vaccine every year.
Derek - That's exactly why we need a vaccine every year. And in many years, the variance of flu that we put in the vaccine, means they have to be changed to track the evolution of the virus.
Chris - You are one of the people who sits on the panel helping to decide what goes into those vaccines. How do you make that decision?
Derek - There's a fantastic worldwide network. I think it's about 150 countries worldwide now, taking throat swabs from people who look like they might have flu and, if it is, checking to see how different it is from earlier variants. This happens in those countries and it also happens by these viruses being sent to one of five laboratories around the world that do very careful analysis. One in Melbourne, Australia, Beijing, China, Tokyo, Japan, Atlanta in the US, and in London here in the UK. Those viruses are tested - as well as seeing how different they are from each other - to test how well our immunity will work against those viruses. And if our current immunity won't work well against the new viruses that are evolving, then we have to update the strain of flu that is in the vaccine.
Chris - What arrives in our winter is going to be six months out of phase with what was doing the rounds in the southern hemisphere. How do you know that what turns up here six months later from Australia is going to be what you think it is?
Derek - Yeah, so this is really the million dollar question because when we make the choice of which strain to put in the vaccine... for example, the strains that people are getting vaccinated with now to protect them against the flu, that will probably come in the next few months. We chose that strain in February this year because enough vaccines had to be made to vaccinate everybody. Currently I think there are about 700 million doses of vaccine made each year. We do the best job that can be done in February to figure out what's going to happen the next year. But it's not a perfect science yet.
... and they expect to be right about 40-70% of the time. But how can we improve on that? Well Simon Fraser University's Jessica Stockdale thinks we might be able to use machine learning to spot patterns in the evolving genetic code of flu that predict the next move the virus will make. As she shows in a paper this week in the journal Science Advances, she's used genetic data from epidemics in previous years to train her system, and then tested it using what we know the virus did more recently to gauge its accuracy. She was hitting the bullseye up to 95% of the time; used alongside existing approaches like those Derek mentioned, it could dramatically reinforce our present vaccination initiative...
Jessica - We take an approach which we can describe as building something like a family tree of our influenza viruses. So we collected about 30,000 public influenza sequences from between 1980 to 2020, and we built this phylogenetic tree, this family tree of those influenza viruses, and used the the structure of that tree, which tracks the mutations that are accruing in flu over time, to look at how the different flu viruses are related to one another to try and predict what was going to grow moving forwards. So our machine learning model looks at which flu viruses are around right now, it calculates various statistics on that tree; how fast are a small family of flu viruses growing? How large is this family? What's the shape to try and classify? Yes or no - will that group of flu viruses be a big problem for us in the next year?
Chris - I guess it's a bit like watching the flu dance floor and seeing what moves it makes, and if you know the dance, you can say, well, when it does that move, it tends to follow it with that move. And so you are doing that for the structure of the virus and therefore you've got some chance of making a guesstimate as to what it is likely to do next.
Jessica - Yes, exactly. It's a great analogy. And it's difficult because flu viruses don't always do the dance moves that we expect that they're going to do. But if we look over the whole globe of all of the different patterns that they're making, there are general patterns that emerge, we might hope, and that's what we're trying to predict.
Chris - Does it work though, Jessica? When you've taught your model using the past, is there not a risk that you end up answering the question that you wanted it to answer rather than what's genuinely going to happen? Can you test it going forward to make sure it's robust?
Jessica - Yes, it's difficult, but that's what we've been trying to do. We did five different experiments on the last five years of our data from 2016 to 2020. We tried to make a prediction, assuming that we did not know what happened. In the end, we masked from ourselves the truth and used that as a method to test our approach. And we found our model to be around 75 to 95% accurate at predicting what was going to be successful the next year.
Chris - The WHO reckon they get it right about 70% of the time, so that would put you at least as good as the current endeavour and possibly better.
Jessica - Yeah. So there's a couple of different ways of measuring success, but we found our approach to be pretty similar to that found by the WHO, despite the fact that we have some more limited data. We're limited to public data only, so that's why we hope that our approach would be a useful addition to the toolbox, I would say, in trying to do this vaccine strain selection.
Chris - So what's the computer doing that we can't with our current WHO panels and so on achieve, or we can but we can't do it as well as your computer system. What's the missing link that you are plugging into?
Jessica - I would say that a human or a person could do what the machine learning model is doing. It's just able to do these calculations a lot faster than us, and there are certainly things that the WHO are taking into account when they pick vaccine strains that we would still want to do, such as, 'is the strain we're selecting even viable for vaccine selection?' But this computer is able to be a helper for us to have this really fast thinking that can add into our own human intelligence that we're using to pick these strengths.
10:21 - AI Safety Summit raises hopes and fears of new tech
AI Safety Summit raises hopes and fears of new tech
Michael Wooldridge, University of Oxford
World leaders and top tech experts have been attending a landmark summit on artificial intelligence at Bletchley Park. The site is synonymous with code-breaking because Alan Turing and his colleagues were based there during the Second World War.
I’ve been speaking to Mike Wooldridge, professor of computer science at the University of Oxford, and a leading authority on artificial intelligence…
Michael - AI has obviously been a thing now for the last decade. We've seen lots of announcements in the press and so on over the last decade. Everybody's got very excited about AI so that over the last year in particular we've seen the mass market adoption of general purpose AI tools for the first time. And of course I'm referring here to ChatGPT. And this has brought the AI safety debate onto the international front burner. And, in particular, there is concern that all of a sudden the rate of progress is such that we need to take AI concerns very, very seriously.
Chris - Who's there?
Michael - We are seeing some world leaders, we're seeing the head of the EU there, Kamala Harris is there. We don't of course have access to the exact list of who was invited, but the big question on everybody's lips was whether China was going to be there. And indeed China has been there. I've just seen pictures this morning of the China delegation speaking. We've got representatives of big tech, most famously Elon Musk, and leaders of big AI companies. So I believe Sam Altman from OpenAI is going to be there, and then there are a number of academic thinkers around AI safety with a number of government representatives and so on. So that's the mix. It's around 120 or so people.
Chris - Rishi Sunak has pushed quite hard on the Terminator style scenarios and the big threat, the existential risk. But there's quite a strong piece which has just come out in the Daily Telegraph by the lady who leads Big Brother watch and she says the elephant in the room here, it's actually a progressive erosion of freedoms. And she uses the fact that we've got police body cams using this sort of technology but in fact we've misidentified twice as many people as we've caught using this sort of technology and it's corrosive for that reason. And she's worried on that basis. And that doesn't seem to be being mentioned.
Michael - This is raising a point that many AI researchers have raised over the last year. And whenever you talk about AIand the future of AI, the conversation always tends to the dystopian very quickly and the most dystopian AI scenario is that somehow AI might eliminate the human race. But at the same time, there are a long list of concerns around the harms that AI can do, for example, around privacy and human rights, around misinformation on a potentially massive scale as we head into elections for which there is really no debate. The communities very largely agree that they are real and they are immediate. Whereas there is a much wider range of views about the existential scenarios. And so I think what this article is doing is pointing to exactly this point. Whenever we start talking about the terminator scenario, it just sucks all the oxygen out of the room. That's the only thing that gets heard and there's never space to discuss these much more immediate and very real concerns.
Chris - They are certainly figuring in the considerations of politicians though, aren't they? Because the EU has got some policies in place on this already, haven't they?
Michael - The EU has had some draft AI regulation now for a couple of years. That's been very widely discussed. There are roughly three models of AI regulation that are in circulation in the world at the moment. There's the US model, which is very innovation led, aims to protect innovation, aims to protect the tech industry. It's fairly lightweight and so on. Then there's the EU regulation, which is very human rights driven, very much out of the model of GDPR which is the data protection regulation that's been with us for a few years, very focused on protecting human rights, protecting individuals by way of identifying high risk scenarios for applications of AI and trying to regulate those. Then there is the Chinese model, which is basically a state-led model, which is very much concerned with protecting the institutions of the Chinese state, the current operation of Chinese society. So those are the three models that are in the air. The UK is in quite an interesting position in that we actually have some freedom, not complete freedom, but some freedom to negotiate our own space there. It's worth bearing in mind though that if we want to do business around AI with the EU, then we're not going to have an option about whether we buy into that regulation. We absolutely will have to buy into it.
Chris - Will this amount to anything? Because it's all very well to say we're going to have some regulations and the UK wants to lead the way in doing this, but there are so many jurisdictions around the world that don't buy into any of this sort of thing, and in fact they trade in fake news; Russia, North Korea. Would they not just potentially continue to be equally or more malignant because of these regulations, even if you put them in place?
Michael - Well, this is one of the many dilemmas. One of the difficulties with AI regulation is there are just so many voices in the room that are just flatly contradicting one another and how you navigate your way through those contradictory voices and contradictory arguments. What do you do if we all play nicely but other countries around the world don't? And let's be completely blunt, obviously nations who are not our allies will be considering how to weaponise it and use it against us: misinformation and disinformation is just one of those. And there are many different ways, and I say cybersecurity attacks is one which is clearly very prominent on the government's agenda, so I do think we have to be realistic about what's achievable.
Chris - We've dwelled very heavily on the negatives and the risks and so on, and it's important that people are cognisant of those. But what are the potential benefits? Where do you see the big wins coming in the next five years from all this?
Michael - I've always been tremendously excited particularly around the applications of AI in healthcare. And I do not see AI as replacing physicians anytime soon, the point is that AI is just going to be another incredibly powerful tool that doctors can use to help them in their jobs and has enormous scope through, for example, wearable technology to give us advance warning of the onset of heart disease, to give us feedback on how stressed we are in our lives. And the long-term benefits of that I think are going to be absolutely enormous. Many people say, well, I don't want an AI doctor, I'd much rather have a human doctor. And that's a rather first world concern. In some parts of the world, it may well be that it's AI healthcare or nothing at all. That's the kind of thing that gets AI researchers out of bed in the morning. That's actually what we're excited about. That's why we're doing what we do because we can see those benefits coming and they are going to be tremendous.
18:16 - Chimps take higher ground for strategic advantage
Chimps take higher ground for strategic advantage
Sylvain Lemoine, University of Cambridge
The history books are full of tales of the importance of using elevated terrain in warfare. But new research from the University of Cambridge suggests it’s not just humans who use hills to their advantage in combat. Chimpanzees in West Africa also use high ground to conduct reconnaissance on rival groups before making forays into enemy territory. The study's lead author is the University of Cambridge’s Sylvain Lemoine.
Sylvain - The chimps when they travel in their territory, they have like these movements toward the outskirts, toward the periphery, which is a dangerous area where they can meet hostile neighbours. So by comparing the usage of elevation in both directions, we found out that they're twice more likely to climb the high hills in the overlap area with other groups. And in the periphery when they move toward the periphery than when they come back.
Will - When they do this, how do we know that this is reconnaissance as opposed to them just enjoying being on top of a hill? Is there a difference in their behaviour that you noticed?
Sylvain - So when they are in the periphery at high grounds, they are more likely to adopt quiet activities like resting where individuals are just not making noise. And combined with the previous effect of being more likely to stop at the high hills on the periphery. While if they were going for other purposes, we should not expect a difference between travelling toward the outskirts of the territory and going back. And we should also see no differences in the activities when we compare the periphery and the core area.
Will - When they are at the top of the hill and they are trying to detect other groups of chimpanzees, what senses are they using? Because it's quite hard to spot something in a jungle, isn't it?
Sylvain - Yeah, so these high spots in the field site where we studied chimpanzees, these slopes are, and the top of the hills are, obviously covered with vegetation. So they are not providing particularly good lookout points, but the acoustic conditions, and so the auditive ability to detect long distance calls, is much more improved.
Will - Say that they're at the top of a hill then and they hear a rival group of chimpanzees fairly close by. How does this affect their behaviour going forward?
Sylvain - When the neighbours are close by, the chimpanzees tend to retreat from the hills when they hear them. But they tend to approach the neighbours when those are much further away. And this effect is particularly prominent when they are on the top of the hills. So that makes us conclude that the hills enable further detection of the neighbours.
Will - Every time there's a new study that comes out on chimpanzees, it's remarkable how similar you can imagine this sort of behaviour being compared to early humans. So we watch them manipulate stone tools and there's even speculation as to how long it'll be before certain chimpanzees can manipulate fire, which is a terrifying prospect. But do you think a study like this kind of gives us a window into the past, into our own human evolution?
Sylvain - Yeah, sure. And that's also one of the main reasons why we study one of our closest living relatives, chimpanzees, is that they give us insight on the behaviour that we have in common. And that means behaviour that could have been present in our last common ancestor. So as research goes we discover more and more similarities. But what it tells us about our own evolution is that there are aspects of intelligence, aspects of collective behaviour that are deeply embedded in our evolutionary past. And that can tell us much more about how our ancient Hominins, and species that have disappeared now, could have thrived in a very competitive environment.
Will - Is there a chance then that this sort of behaviour, seeking out high altitude areas for reconnaissance and protection, might have been what put us ahead of the pack when it comes to the rest of hominids?
Sylvain - This is a possibility. There is a body of theories called the Complex Topography Hypothesis that states that when the transition between forest and savannah dwelling, ancient hominids could have used cliffs and plateaus, typical landscape found in East Africa, to thrive in a changing environment. And that could have helped them in hunting strategies but also anti predation strategies, themselves avoiding being eaten by big cats. So then remains another possibility of the usage of this tectonic landscape, which would've been in competition with other groups of the same species or even competition with other human-like creatures in that time. So that remains a possibility. And the importance of high ground in military tactics seen nowadays tells us as well that we have probably kept certain fundamentals of this collective action that take place during warfare.
23:05 - Off you pop: Is cockney on its way out?
Off you pop: Is cockney on its way out?
Amanda Cole, University of Essex
New research from the University of Essex suggests that the cockney way of talking is disappearing, and new accents are beginning to dominate. Our own James Tytko, speaking one of these - Standard Southern British English or SSBE, an updated version of the way the King speaks apparently, has put together this report…
James - Language is constantly and quickly evolving. Incorporating new influences and sidelining old ones. Once spoken by people of all ages in this corner of the world, accents like cockney and Queen's English were not represented in a recent survey of 193 young people between the ages of 18 and 33 from across Southeast England and London. It seems they will soon be consigned to history. Using a computer algorithm, researchers at the University of Essex split the participants of their study into just three distinct accents - 'Estuary English,' Standard Southern British English, And last but not least, multicultural London English. How do we objectively define and categorise someone's accent and why does it matter anyway? I spoke with Amanda Cole who led the study to give us the answers.
Amanda - So some of the defining features of estuary English are quite similar to Cockney. So for instance, we found that young people, when they said words like mouth or bout, they had their tongue a bit further forward in the mouth. So it might sound more like maths or bat or Soufend. So that sort of feature. This is a feature that we find a lot in cockney, but we would find it to a greater degree in cockney that vowel would be a bit more pronounced. For standard southern British English, essentially what we're finding is something that does quite closely resemble what we would call 'receive pronunciation,' so RP, but what many people might refer to as say Queen's English or BBC English. SSPE does have some similarities with that. But again, we could kind of consider it a slightly more muted version of that. Obviously that's not a linguistically accurate term in a way to call it more muted. But if we can kind of picture it as this less extreme version of Queen's English that's closer to more regional pronunciations. Multicultural London English is a variety that has emerged first in London. It was first documented in East London and it was thought to have gone back to around the early 1980s. And it has lots of new and interesting and exciting linguistic features that are different to what we would find in other accents around the southeast.
James - You mentioned that the estuary and standard Southern English accents might be described as being less pronounced perhaps than their cockney or Queen's English ancestors. Do you think we're seeing a sort of gradual erosion of regional accents?
Amanda - I think part of what we're seeing is dialect levelling across the sort of region or a geographic space. People begin to speak more similar to each other than they would've done in previous years.
James - Dialect levelling is thought to occur as a result of the greater distances our voices travel these days, both physically and virtually, which results in greater contact and therefore integration between speaking traditions. But this is something of a two-way street.
Amanda - We can talk about dialect levelling, but that doesn't mean that new and innovative ways won't emerge. So multicultural London English is an example of that where it kind of bucks this trend of levelling.
James - In Britain, especially, I think the way you speak has a bearing on how you are perceived by some people. And was that something you were interested in, in this study into accents in the Southeast?
Amanda - Yeah, definitely. The fact that we're saying that maybe cockney and Queen's English aren't as common as they were isn't a problem. It's okay for accents to change. This can kind of lead to these ideas that English is going to the dogs, that English is decaying, that people aren't speaking it correctly anymore. And these ideas are all false. There's no inherently correct way of speaking. There's no logical or scientific way that a person can configure their mouth and make sounds that are inherently any better than any other way of speaking. That's all a social construct. The way that we talk reflects who we are. It reflects our class, it reflects our ethnicity, it reflects where we're from. And we shouldn't expect anyone to have to kind of leave that at the door and speak in a certain way. We should be promoting and accepting linguistic diversity.
29:17 - Can you bend a laser around the Moon?
Can you bend a laser around the Moon?
Thanks to Professor Michalis Zervas and Professor Simon Hooker for the answer!
In the far flung future, when we need to send messages out into the cosmos, we could do worse than using lasers. According to NASA, future Laser communications will enable 10 to 100 times more data transmitted back to Earth than current radio frequency systems. But what if there’s something in the way? Well perhaps gravity can help us out. Here to explain is University of Southampton’s Professor Michalis Zervas…
On a cosmic scale, it is known that light and, therefore, a laser beam can be deflected or bent around a celestial body, like the moon, due to a fascinating effect called gravitational lensing. Massive objects, such as a stars or galaxies, warp the spacetime around them, and this warping causes the path of light to curve as it passes near the celestial object.
However, compared to stars and galaxies, the moon has much, much smaller mass and, as a consequence, creates a very slight gravitational lensing effect. Therefore, its impact on laser beam bending, although theoretically present, it will be extremely subtle and typically not observable with current technologies.
So, although it will be possible to deflect a high-power laser beam around and beyond a big star or a galaxy, it cannot be done around the Moon.
Gravity is a no go for the Moon then, but what about man made stuff like mirrors and special beams? To talk us through that, the University of Oxfords Professor Simon Hooker...
Setting aside the problem of getting the mirrors in the right place, and keeping them there, you’d find that they’d need to be very large. This is because the laser beam will naturally expand on its way to the moon due to a process known as diffraction. Perhaps paradoxically, first expanding the laser on Earth helps, but even if you launched a beam with a diameter of 1 m, it would expand to nearly 200 m on arriving at the moon. So you’d need to put some pretty big mirrors in space!
Something else you might try is a special kind of optical beam known as a “Bessel” beam. This can be made by passing a laser beam through a glass cone, which converts the beam to a conical beam. We can think of a conical beam as being made up of lots of beams all heading from points around the rim of a circle to the axis of the glass cone. These beams all overlap at points along the axis to give a very bright central spot, that looks just like a focused laser beam. Bessel beams have “self-healing” properties. This means that if an obstacle is placed in its way, the central bright spot will be blocked. But, all the laser light away from the axis will go round the object, and this can “re-build” a bright central spot at points downstream. So, it looks just as if the beam has re-built itself —or self-healed!
Could we use this self-healing property to project a beam “through” the moon? In principle yes, but at the moon the diameter of the conical beam would need to be many times that of the moon (so that the light going round the moon can re-build the Bessel focus beyond the moon). On Earth, the beam would need to be bigger still … which would be expensive and unwieldy. So, on balance, I think that it would be easier to wait a few days for the moon to move out of the way!