The marine carbon cycle, teaching robots, and a mental health zine

Gareth Mitchell:               Hello, everyone. I'm Gareth Mitchell. Welcome along. Today, the latest from Imperial's REACT study. That's in just a moment. Also, how our oceans soak up CO2 and how fishing isn't helping. Not only that, but I'm learning how to train a robot to make toast.

Dr Edward John...:           So I'm going to move the robot now. And I'm guiding its fingers down onto the lever of the toaster. Okay. So now I've kind of taught the robot the nature of the task.

Gareth Mitchell:               And we dip into a mental health zine. Or is it zine? Stay with us.

Gareth Mitchell:               Right, folks, say welcome along. And we're going to start with some news. And it'll be good, I think, to jump into the latest from the REACT study, if we may, with Justine Alford. Well, obvious question, what is the latest then from the study?

Justine Alford:                   Hi. Yeah. Thank you very much for having me on. So the latest from the REACT One program, which stands for the Real Time Assessment of Community Transmission, which is looking at coronavirus prevalence in England. There's a mix of sort of good news and slightly less good news in this latest announcement.

                                             So what they found is, compared with December, that infection rates have gone up a huge amount. So they're actually three times higher than the infection rate was in December. However, from early January to mid-January, infections had actually decreased quite substantially. So there is some good news in there as well. But in the most recent data, it does actually look like that trend has begun to stall and the infections have been flattening.

Gareth Mitchell:               What does that suggest to us then?

Justine Alford:                   So this could be due to a number of factors. It's not actually possible to tell from this specific data set, why these different trends are happening, but it could be due to a variety of reasons. It could be that people are behaving differently. It could be because children have gone back to school since the Christmas holiday and are mixing more. So yeah, it could be a variety of reasons that now the infections seem to be plateauing.

Gareth Mitchell:               So infections appear to be plateauing. What, if any, are the causes for concern? What should we still be worried about Justine?

Justine Alford:                   Well, infections might be stalling at the moment, but they appear to be doing so at a very high rate. So a huge number of people are still carrying the virus around England. So it's about one in 23 people had the virus during this most recent testing period. So of course there is concern that that could lead to people going to hospital because of their infection. And ultimately it could cause some people to unfortunately die from their infection. So there is concern that while infections don't appear to be going up at the moment, they are still at a very high rate.

Gareth Mitchell:               All right, Justine, thank you very much indeed for that Justine offered there. Right, so Hayley Dunning joins us now, and this is my word of the day, definitely. There's a holobiont, and forgive my ignorance, Hayley, I'd never heard this word before. I don't know what it means. So be gentle with me. Just break me into this story then, what's the holobiont and why are we talking about it?

Hayley Dunning:               Well, I could pretend to be the all-knowing person here and say, I always knew what a holobiont is, but I didn't either until I talked to the researcher. So what a holobiont is is a term given to large organism, say a plant or an animal or a human, and its associated community of microbes. So this is often called the microbiome. And a holobiont is considering these two things as kind of one entity. So you can imagine how our microbiome in our gut helps us with the good bacteria and keeps us healthy. And that's part of how we exist as humans. So it's considering these all together. And it's also the subject of a new center, primarily at Imperial. It's just been given 10 million pounds of funding by the Leverhulme trust, to research into holobionts with some other universities and some other institutions, but based Imperial.

Gareth Mitchell:               What will the center actually do?

Hayley Dunning:               So first we need to know what holobionts are out there. So they want to create a holobiont tree of life, really looking at how organisms and microbes are related across all of life. And from this, obviously, when we've got this basic understanding, we can now take it to answer some practical questions in things like biodiversity and agriculture.

                                             So in the biodiversity realm, there's things like animals in zoos. Their microbiomes may have been changed by living in zoos. And if we want to reintroduce them to the wild, we might have to know what their natural microbiomes are like in order to help them adapt back into that situation. There are also some threats that we know are associated with microbiomes. For example, coral bleaching is triggered by high temperatures that cause the essential microbes to disassociate from the coral host. And that obviously disrupts the whole coral reef ecosystem. Now teams Imperial looking at this area already, we have a few. So there's some that are looking at the impact of pesticides on the holobiont of bees. And also how the skin microbes of some amphibians help them fight off a deadly amphibian fungus.

Gareth Mitchell:               Well, it certainly sounds as if the new center has its work cut out. So just briefly then anything else that's on the agenda here, Hayley?

Hayley Dunning:               Yes. So the main other strand they're looking at is agriculture. As I mentioned, plants can also come associated with microbes. And these are often in the roots and they help the plants take in lots of nutrients. But this is not always the most efficient process. So they're actually looking at how they could understand and then perhaps bioengineer crops and their microbes together, considering them as a holobiont to make them more efficient.

Gareth Mitchell:               Thank you very much indeed for that Hayley. Now good news for you, dear podcast listener, if you're a Hayley fan, because there's even more of Hayley coming up now in this interview. We're going to talk about the environmental effects of fishing in the oceans. And this isn't just a story about over-fishing. It's about how feeding ourselves can compromise an important function of the ocean. The seas act as a giant carbon sink, capturing the greenhouse gas and blocking it away. Tiny single cell plants and algae photosynthesize CO2 out of the air and then sink down to the sediments below. Well, Hayley has been hearing about new research in the journal Global Change Biology that says fishing disrupts this cycle. Hayley caught up with study authors Dr Emma Cavan of the department of life sciences at Imperial and Dr Simeon Hill of the British Antarctic survey.

Dr Simeon Hill:                  Fishing is essentially an operation that removes large amounts of biomass from the seas. This biomass is living organisms that actually perform valuable functions within the seas. And one of those functions is to play a role in these carbon cycles. So what actually happens is that fish and other Marine organisms eat food. This food is made out of phytoplankton. Some of this food is used for growth and maintenance in organisms like fish. And some of it is egested, passed out of the body as feces.

                                             And that feces sinks to the bottom of the ocean. And if it reaches the sea bed, it has a high chance of keeping carbon dioxide out of the atmosphere for a long period of time. So simply by removing biomass of fish, fishing could be disrupting this process, reducing its efficiency, reducing the amount of carbon that's being stored on the sea bed. But fishing also disturbs Marine ecosystems in various ways. And what we did in our study was to explore the potential roots of disturbance.

Hayley Dunning:               So what exactly did you do in this study? What did you look at?

Dr Emma Cavan:              We explored where there could be a spatial overlap of fishing practices and also this plankton carbon sink. So we used freely available satellite global data of fishing intensity. We also got a global picture of what the carbon sink is like. And what we found is that, as we expected, both fishing and this plankton carbon sink are really high near the coasts and on kind of shallow continental shelves. So particularly around the UK and Europe.

                                             And we expected to find this because both fishing and the carbon sink are dependent on plankton. And so this wasn't kind of a "wow, we found this crazy, amazing new thing that we never expected." It was really just highlighting that these two really important ecosystem services that are important to humans, fish for food and carbon to maintain a natural carbon cycle, are happening in the same place. And so any kind of negative impacts of fishing could be really damaging to this carbon sink.

Hayley Dunning:               So what needs to happen next now that we've identified this area, is a case of going there and trying to figure out exactly what impact fishing has.

Dr Simeon Hill:                  Our study shows that there's a great deal of overlap between these two ecosystem services. But what we haven't been able to do at present is to quantify the extent of the impact of fishing. So this really is the next step is to identify, does this overlap lead to significant impacts on the carbon cycle.

Dr Emma Cavan:              Certainly a lot of research needs to happen in this space. And also there's a lot of policy needs for this information. So we've already had different NGOs, charity organizations, as well as policy makers, both in the Europe and in the UK, being in contact to ask this information. And so we need to work with the scientists and policy to really get this information.

Hayley Dunning:               If fishing is shown to be significantly disrupting the ocean's ability to take carbon from the atmosphere, what can be done about it?

Dr Simeon Hill:                  At the moment, we do need more information to quantify the extent of this problem. But there are some fairly obvious solutions. One of which has been to rebuild over-exploited stocks. So that's where fishing has driven bio mass of fish species down to very low levels. We need to allow them to recover. Another possible outcome of this type of research is to highlight hot spots of carbon sequestration in the ocean and afford them some special protected status.

Dr Emma Cavan:              The kind of policies that could be in place are things like changing fishing practices. So moving away from things like bottom trawling, and that's where huge, weighted nets are trawled along the sea floor. And that re-suspends a lot of the carbon. So that's another way that fishing can be really damaging, not just removing biomass, but also disturbing carbon sinks and sediments that are rich in carbon.

                                             And also there's things called MPAs, which are Marine Protected Areas. And on paper it can be seen as we have X percentage of our seas, territorial waters are protected. But actually patrolling that and making sure there's no fishing is much harder. And so we still see a lot of over-fishing happening in these supposedly Marine Protected Areas. And so now NGOs are really excited about this research because they can say, right, if protecting Marine life from over-fishing and protecting dolphins of bycatch, isn't enough of incentive to make these policies actually mean anything. Perhaps reaching net zero and leaving carbon locked in these sediments will be a good motive for politicians to do it.

Gareth Mitchell:               Emma Cavan and Simeon Hill. And that research is published in Global Change Biology. Well, I've been to meet a robot, one that learns quickly and easily. Now this isn't a bot that looks humanoid. It's a robot arm. And if you need to train it to perform a task, you just show it how by grabbing the arm whilst you do the movement yourself. Next time round, the robot does the same thing on its own. Now, to be honest, I'd never really thought how you train machines, but basically it involves a whole load of programming. So a simple human-friendly way of teaching bots how to do things, well, that's a big deal. As I've been finding out at Imperial's robot learning lab. Which presented its latest work at a robot learning conference just before Christmas. The lab's director is Dr. Edward Johns. And when I drop in, we're teaching the bot how to push down the lever of a toaster.

Dr Edward John...:           So I'm going to move the robot now and I'm guiding its fingers down onto the lever of the toaster. Okay. So now I've kind of taught the robot the nature of the task. And the robot's moving around the toaster and it's got a camera on the wrist. So the robot's moving in and out of the toaster now and collecting images of what that toaster looks like from lots of different viewpoints. And with this data, what it's doing is it's training a neural network to map from the images to where the robot should move to.

Gareth Mitchell:               So you've shown it what the gesture is, where it's just bringing down it's robot arm onto the toaster lever as if it's putting some toast down. And then in this second phase where we can hear the robot whirling around and the arms are moving around, hovering above the toaster, it's then teaching itself in a way what the toaster looks like. So you've given it the gesture and now it needs to figure out what a toaster looks like.

Dr Edward John...:           That's right, yeah. So an alternative way to do this would be for you to, as the engineer, sort of come up with an algorithm which allows the robot to find the toaster. But of course that would mean that you would have to come up with some specific code that's tailored towards just toasters. And that's not very scalable because of course, everyday people don't know how to write code to recognize toasters. So we want the robot to be able to learn by itself what the toaster looks like by automatically moving the camera around the toaster and then building up these data set of images, which can then be used to train the neural network.

Gareth Mitchell:               And as you were saying that on the screen, on the computer now, a rather grainy image of the toaster's appeared. So this is the robot's eye view of the toaster. This is what it thinks it looks like.

Dr Edward John...:           Absolutely, yeah. So it's quite a low resolution image. And it's not even obvious, if you were to look at that, and I said, what is that image? You wouldn't necessarily know that's a toaster. And also the robot doesn't know it's a toaster. All it knows is that it should move towards some part of that image because that's what I've told it from the demonstration. So I can show you that happening now. Okay. So the robot's now moving into the toaster, aligning itself with the images that it's captured. And then moving the lever down.

Gareth Mitchell:               Whoa.

Dr Edward John...:           You can see the robot operating autonomously.

Gareth Mitchell:               It brings the arm down absolutely flawlessly onto quite a narrow little handle. I guess the real test now is you've just moved the toaster. Can it still do it?

Dr Edward John...:           So that's really what I'm trying to do here with moving the camera around the toaster, so the robot understands what it looks like from different viewpoints. So now what we are really doing is testing its ability to generalize to different positions of the toaster on the table.

Gareth Mitchell:               So the toaster's been moved. Can the bots figure out how to press the lever down to toast the toast? It's going in.

Dr Edward John...:           So there we go. Successfully toasted some virtual toast that isn't there right now, but it will be one day.

Gareth Mitchell:               But it has proven. There you go. So what you've proven here is that you show the robot a particular movement by forcing its hand, as it were, to make the movement. As if you were with a child, you say, come on, you take the child's hand and you press the toaster like this. That's what you started doing with the robot. Then it did its own map of the toaster so it has a sense of how the toaster, as a form, contrasts with the blue tablecloth. But then what really just what I thought was amazing, is you move the toaster around and the robot can figure out that this thing that you and I would call a toaster has moved. But it repeats the same gesture on the same part of the toaster. It didn't just try and randomly press right into the slots where the toast goes, for instance.

Dr Edward John...:           Yeah. So we can move the toaster anywhere on the table and it's going to cause the image of that toaster to then change. And then the robot knows where the toaster is relative to the camera, and therefore the robot can actually move. The demo just provided you was the robot operating the lever on the toaster. But of course before that, the robot would need to place a piece of bread into the toaster. And so that's what is called a multi-stage task.

                                             So one of my students, Norman Dipallo, did some research on this. And he actually had to do this in Italy during the lockdown. So he actually built his own robotics lab at home and bought a very cheap robot arm, set it up at home and then do all of his experiments at home just by himself. And I managed to publish a paper on that research. If you work in robotics, you really, really need to get hands on with the robots. And there's only so much you can do just kind of thinking about algorithms and doing simulations. Until you really work with physical robots, then you don't... You often kind of overlook some of the real challenges that we have to face.

Gareth Mitchell:               So where's this work going next? Is there a next for this or have you cracked it now?

Dr Edward John...:           Well, I think that, yeah, the next would be dealing with much more complex tasks. And also figuring out how we can actually scale this up properly. So let's say we had a thousand people all teaching robots across hundreds of different tasks. How can we distill all of that information down to one neural network? One that all the robots can then use together. Because at the moment we're just kind of training individual tasks. So what we'd like to do is to be able to train multiple tasks and allow all of that information to be shared and for robots to be able to learn perhaps on one tasking for that information to be able to help the robot learn a new task very quickly.

Gareth Mitchell:               Does the robot have a name? Or a nickname?

Dr Edward John...:           No, not really. It's called Sawyer, which is the manufacturer's name. But we don't really have our own name. Just robot. Because we only have one. When we get another robot, maybe we'll have to start giving them different names so that they feel a little bit more loved.

Gareth Mitchell:               Right. Okay. Well, Edward and robot. Thank you very much.

Dr Edward John...:           Thanks very much, Gareth.

Gareth Mitchell:               Edward Johns and the robot with no name. Well finally here's Justine.

Justine Alford:                   Hi, my name's Justine Alford and I'm from the Institute of Global Health Innovation. And today I'm here with IGHI researcher, Dr. Lindsay Dewa, who's going to talk to us about a new mental health theme that she's launched. So first things first, Lindsay, could you start by telling us what exactly a zine is?

Dr Lindsay Dewa:             Of course. A zine, or a zine is how we've been describing it, is a small magazine essentially, in both style and content. So we've really high quality images. And I guess for us it was a research zine to showcase our research in an easily accessible and engaging format for young people.

Justine Alford:                   Fab. And so where did the idea for this zine come from?

Dr Lindsay Dewa:             We've been working with our public partners, the Mine Map, and we actually considered and actually completed a zine or a zine. Well actually, it was a glossy, attractive sort of focus zine. So not so much a research zine. But that was to increase awareness of the needs of young men and their mental health. I guess we took that concept and thought, why not do the same for our current project, looking at digital social connection? And actually the young people loved that idea that we were working with on the project. And we thought it was a really great way to of get the research out to young people, both in a print format and an online format. So it was really accessible to everyone.

                                             And actually we thought, why not make it more of a regular thing? So not just for the Quality Social Connection project, but actually for any project that involves young people. Why not also, in addition to the academic papers, et cetera, but have a much more public facing document that they can access and really understand what we do here at IGHI.

Justine Alford:                   Great. And as I understand this zine or zine was co-produced with young people who have experience of mental health difficulties. So how did you involve young people in creating it? And why did you feel that this was an important thing to do?

Dr Lindsay Dewa:             For the Quality Social Connection project in digital interventions, we'd already worked with a nine person, very diverse, young person group throughout the whole research stages. And the last stage, of course, was dissemination about how we were sharing our research out to the young people and other audiences. And so we'd already co-produced a video and an infographic. And then of course we was going to do the zine as well. So we had a two and a half hour workshop with a young person group, and we worked with two designers as part of the Helix center. Whilst the young people were informing what the zine was going to look like, what the format was like, what the structure, what we should include.

                                             The designers were doing an initial sort of scoping exercise, like on the screen, to see what it could look like physically, at the end. And so it's very much informed by the young people. And they thought it should be colorful, engaging, simple language. And so that was sort of the initial step. And then we worked together then. I took those, all of those concepts and ideas and decisions into some software to bring all that to life, and then worked over a series of weeks back and forth with the young people. They had access to the software, they were able to shift and shape as well.

                                             And then, yeah, we did that over a period of time. And one idea actually I specifically wanted to mention was about their idea about interviewing a person that was in the more public eye that was more like a celebrity or a more well known figure. And so a mental health ambassador, for example. So one of our young people, Nathan, interviewed Johnny Benjamin, who was a mental health ambassador. And that worked really, really well. And that's one of our main features of our zine. As well as an interview with a young person.

                                             So Ellie, one of our young people was interviewed about her experience in the project as well. And so it's really important to involve young people in this because nothing about them without them is a philosophy. And that's so, so important. So essentially we're trying to include them in everything, we work together throughout all stages. It's sort of an ethical consideration in everything that I do and really, they just make it so much more impactful, much, much better outputs. It's informed. It's decided by them. And it just makes for a much better research project that actually is about them, for them.

Justine Alford:                   Yeah, of course. And it sounds like it makes it much more close to their hearts and much more something they will actually want to read and engage with. So ultimately, what are your hopes for this zine? What impact do you hope that it will have on the mental health sphere?

Dr Lindsay Dewa:             So obviously that's the first issue looking at digital social connection as a feature. And the next project we're looking at climate cares. So specifically about climate change and mental health in young people. And featuring the changing Wales project, which is actually looking at those relationships with COVID as well. And we are going to be talking to Emma Lawrence, who is leading that project. And I'm expecting an ambassador of the climate cares team as well to be featured as part of that project. And I expect the next one will be around March, April time, which will be featuring that project.

Justine Alford:                   Fantastic. We really look forward to reading about that and you can already access the first edition of the zine on IGHI's website, if you head to the mental health section. So thank you very much, Lindsay, it was really great to hear from you and we look forward to reading the next edition.

Gareth Mitchell:               Yes we do. That's Lindsay Dewa speaking to Justine Alford. And on this podcast, we'd better start working on our next edition. They soon come round you know. That's it for this edition though. Do check us out on iTunes, Spotify, YouTube, and beyond. Thanks to Justine, Hayley and everyone else who's been on today, including the robot. And thanks to you for being there. I hope your January has been panning out okay so far, and I'll see you in February. I'm Gareth Mitchell. Bye-bye.