Is AI watching you? The hidden links between research and surveillance

0:00Therapy's biggest impact doesn't show up in some dramatic breakthrough. It shows up in the smallest moments. You bite your tongue instead of escalating. You put your phone down instead of firing off that text. You say what you actually mean instead of what would hurt someone else the most. Nobody applauds those moments, but they quietly change your life. That's what growing feels like, and that's exactly what Grow Therapy is here for. Grow is therapy that works, covered by insurance. Sessions average$21 with insurance and some pay as little as$0 depending on their plan. There are no subscriptions, no long-term commitments.

0:37You just pay per session. Grow connects you with thousands of high-quality licensed therapists across the U.S. offering both virtual and in-person sessions nights and weekends. You can search by what matters, like insurance, specialty, identity, or availability. And you can book a therapy session in minutes and begin in as little as two days, not the national 48-day average. See if you could pay as little as$0 for therapy in seconds at growtherapy.com slash ACAST. That's growtherapy.com slash ACAST. growtherapy.com slash ACAST. Availability and coverage vary by state and insurance plan. Grainger knows.

1:19If you're a procurement manager at an industrial manufacturing facility, you're balancing inventory and budget while sourcing critical supplies across the operation. Too much inventory? Costs climb. Too little? Production slows. Grainger's$2 billion in on-hand inventory helps you get what you need when you need it. So you can help keep budgets balanced and lines moving. Call 1-800-GRAINGER, click grainger.com or just stop by. Grainger, for the ones who get it done.

2:13Welcome back to the Nature Podcast this week. The link between AI and surveillance. And the first images from the Vera C. Rubin Observatory. I'm Benjamin Thompson. And I'm Sharmini Bundel.

2:35Is AI research being co-opted to keep track of people? Nick Petrichow is here with the story. Many researchers, policy makers and activists argue that AI research could be used to power mass surveillance, and want more critical discussion around how this technology is being used. Their argument makes logical sense. Computer vision, the area of AI dedicated to getting machines to see, can be used in self-driving cars to detect humans crossing the road, or used to automatically tag someone in a social media post. So it doesn't seem like a huge stretch to imagine that it could also be used to track people going around their daily lives.

3:23But this argument has been contentious. If you ask various people about what computer vision does, various people will give you different answers. This is Abeba Bahani, an interdisciplinary researcher who's been looking into this topic. Surveillance studies people will tell you computer vision is doing surveillance, whereas computer vision researchers themselves might hold on to the idea a given technology can be used for good or bad. On the one hand, computer vision can and is used for many useful things, like detection of problems on medical images, biometric security and even for tracking players and balls moving on a football pitch.

4:08So how much computer vision research could be used in surveillance technology? It's a question that Abeba and her colleagues wanted to answer. What we wanted to do with the study is to support these various claims with large-scale empirical evidence. That study is in this week's Nature. To determine whether computer vision research could be used to power surveillance, Abeba and a team of researchers looked at papers published in the Conference on Computer Vision and Pattern Recognition. This is one of the leading academic publications on this topic. To start, the team manually searched through 100 randomly selected computer vision papers and another 100 randomly selected patents linked to these papers.

4:59Specifically, they were looking for words related to surveillance. And based on this scheme that we developed, we read through every line to kind of extract and thematise what the authors themselves claim that these technologies are doing. And we developed keywords that are associated with surveillance or surveillance enabling keywords. They ended up with 30 keywords that were related to surveillance as they defined it. The gathering, extracting or attending to data connectable to other persons, whether individuals or groups. Those keywords included age, anatomy, ethnicity, face, gender, military and crime With this list, they then scoured through more than 19 ,000 other papers published in the Conference on Computer Vision and Pattern Recognition and more than 22 ,000 resultant patents We found that an overwhelming amount of uses of both papers and patents is in surveillance In fact, the team found that 90 % of papers and 86 % of downstream patents extracted data relating to humans.

6:18Such data could be used to develop technologies for surveillance. And 71 % of papers and 65 % of patents explicitly extracted data about human bodies and human body parts. and just 1 % of papers and 1 % of patents didn't look at humans at all. The focus on extracting human data appears to have increased over time too as the thousands of studies the team looked at covered research from 1990 to 2016 and according to their analysis only 53 % of patents in the 90s could enable surveillance compared to 78 % in the 2010s. Whilst could enable doesn't necessarily mean it is used for surveillance, Abeba worries that this technology could harm democratic freedoms.

7:14This has implications for freedom of speech, freedom of movement. If you look at, you know, governments, you know, including from the Western world to African nations, you see that these governments are utilising these surveillance systems to crush down dissent, to identify any leaders in a riot or in a demonstration, so basically to deprive people of their rights to freedom and freedom of speech. The team also uncovered that often the fact that this research could be used for surveillance has been obfuscated. Sometimes you would read an entire paper or patent and have no clue or there is no information showing that what type of data is shared or collected.

8:06Then you look at the figures and you see that it's humans. Instead of referring to humans, some papers would instead call them objects. And the places where they live? Scenes. This is something that Jafin Sadowski, a computer researcher who's been writing a News and Views article on this paper, found particularly worrying. It abstracts away any of these potential issues related to critical reflection and inquiry. What's the actual purpose and use of the models that I'm creating? Do I agree with that purpose? And with the interest of the people who are using them, It abstracts away questions of ethical responsibility or political controversy.

8:51What could be controversial about training a machine vision model to recognise objects? Overall, Jafin was impressed at the scale and rigorousness of the new study in Nature. And he shares Abeba's concerns about the potential impacts that much of the computer vision literature being linked to surveillance could have on democratic freedoms. He also notes that many massive corporations, military and policing operations have a keen interest in the development of these systems. I think that the influence here is that it's the companies and government agencies that are interested in having high tech capabilities for visual surveillance.

9:42Those are the people that are influencing the trends of research and development in this field. In fact, the US military has invested hundreds of millions of dollars into this technology. and governments too have been big investors. Facial recognition cameras are being rolled out for the police and border control use as well. This technology could be used to catch criminals, find lost children and secure borders, so some will find such usage reassuring, but others view the use of this technology as a violation of human rights and freedoms. Abeba though hopes that this new study will help shine a light on how computer vision research is being used, something that could help empower activists, researchers and governments who want to protect their people.

10:36Most of the general public tends to have very little autonomy, very little say and very little option, very little leeway where we have no input in what kind of surveillance technologies are built, what they are used for. Maybe if there is some kind of way to balance, to give people more autonomy, more option, then we are looking at a more level playing field where we can, you know, sit down and look at what are the trade-offs, the positives and negatives. That was Abba Bahani from Trinity College Dublin in Ireland. You also heard from Jathan Sadowski from Monash University in Australia. For more on that story, including a video and an editorial, check out the links in the show notes.

11:20Coming up, how the Vera C. Rubin Observatory will help astronomers map the universe. Right now though, it's time for the Research Highlights with Dan Fox.

11:33Researchers have developed a tiny robot that can manipulate minuscule drops of liquid. Scientists in a range of fields routinely need to move, split or merge drops of liquid. Magnetic techniques can be used to do this, but these methods typically have either limited capabilities or a tendency to contaminate the droplets. To address this, a team developed a droplet-manipulating robot, about the size of a pencil-top eraser. To do this, they incorporated magnetic particles with sugar crystals in a flexible, soft polymer. Dissolving the sugar left holes, increasing the surface area of the polymer and its ability to adhere to droplets.

12:15Finally, they modified the surface so that the tiny robot would attract droplets, while being remotely controlled using magnetic fields. The researchers demonstrated various applications, including bringing droplets together, in some cases starting a chemical reaction, and splitting large droplets into smaller ones. This robot can control droplets at a microliter scale, but a future version might be able to manage nanoliter ones. Read that research in full in nanotechnology and precision engineering.

12:50Newly installed sensors have pinpointed a hydrothermal explosion at Yellowstone National Park in the US. Hydrofermal explosions are violent eruptions of steam and rock that occur when hot liquid groundwater suddenly turns into gas. Volcanologists think that many such events remain undetected because they occur in remote areas or at times when people are not present. Researchers analysed data from an array of sensors that were installed in 2023 to record weather information, seismic waves and infrasound waves, which are low-frequency acoustic signals. The researchers used this data to pinpoint an explosion to a precise time on 15 April 2024.

13:34The event was preceded by several years of increased activity of hydrofermal water discharges in the area. Collecting data on hydrofermal explosions could help researchers to learn to read the warning signs and prevent harm to visitors. You don't need a sensor array to pinpoint that research. It's published in Geophysical Research Letters.

14:02This week saw the first images released from the Vera C. Rubin Observatory, named after the late American astronomer Vera Rubin, who pioneered the study of dark matter. A significant new tool for astronomers is always exciting, and although podcasting is primarily an audio medium, we're going to talk about the images and a little bit about the observatory. Joining me to do so is Nature's Davide Castelvecchi, who covers all things space. Davide, thank you so much for joining me. Hi, Ben. Thanks for having me. So tell me a little bit about this observatory. It's on a mountaintop in Chile, right?

14:38Yes. It's not in one of the really high telescope locations in Chile. Like, it's not 5 ,000 meters, it's only like 2 ,000 something, but it's in a region where you get very clear skies. And making it was a big undertaking. Yeah, it was built by extraordinary cooperation, collaboration between the two largest physical sciences funders in the United States, which are the Department of Energy and the National Science Foundation. The total costs are up to more than$800 million, and it will start its regular operations and its exhaustive serving of the southern sky in the next few months. And key to this is an absolutely enormous digital camera.

15:19I've read that it's the size of a small car, something like that. Yeah, so the camera has like unprecedented ability to capture data and sensitivity. But it's also just the whole way that the telescope was designed to capture a lot of light and to move quickly from one location in the sky to another so that it takes a shot. And then 40 seconds later, it's already taking a new shot. And, you know, moving something that weighs tens of tons is not easy because things start shaking and then it starts and stops and then it's going to be, you know, wobbling for a while. So they really had to design the whole thing so that it would be stable.

15:57And you mentioned the southern sky there. What is this setup going to be used for? The original idea for building such a telescope was to basically survey the distant universe. and how galaxies are distributed and how what we see of galaxies is affected by dark matter, which constitutes a lot of the mass of the universe, actually more mass than the regular matter. But as the idea was being developed, researchers realized that having such a facility would provide opportunities for doing all sorts of other research. So there's the surveying of the distant universe, but there's also a number of other really groundbreaking applications.

16:41So, for example, seeing the sky change from one night to the next and seeing if there are stars exploding, stars acting erratically, and even closer to home, they will be able to cover the entire sky that's available to them every three to four nights. Oh, wow. So, for example, they'll be able to see enormous numbers of asteroids and classify the asteroids that are potentially dangerous in terms of hitting Earth. Well, let's talk about the images that have been released in this first tranche then. They were taken in April, as I understand, during a trial that began when construction of the telescope was completed.

17:18And they're quite something. Listeners, head over to the show notes for a link to Davide's story where you can see one of them. And one of these photos is looking at a particularly dense part of the Milky Way. Yes, in fact, one of the first places they pointed the telescope at was the region of the central Milky Way galaxy, which is known as a stellar nursery. It's the Trifid and Lagoon nebulas, which are dense with ionized hydrogen and where clouds of gas are coalescing into stars. And there's a lot of stars in the process of forming or just young stars that have formed recently. And were these images sort of showing off this is what the observatory could do?

17:56Or is there any science in there that researchers are having a look at? The images that we discussed in our coverage were selected primarily to showcase the capabilities of telescope and to, shall I say, inspire. Together with those images, they also released what they call data products, which are the actual catalogs of objects and measurements of luminosities of objects that astronomers can use. But one really amazing thing is that although none of these images was taken with a single shot, all of them were composed from multiple shots, it still took a surprisingly short amount of time to take in all the data.

18:35So for example, the nebulas that I mentioned were observed in 678 separate exposures. Each of them was monochromatic because their camera is monochromatic and they use filters to separate the different bands of light. But the whole thing took just seven hours of observing time. Now, when something like the Hubble or JWST do these composite images, first of all, they cannot do such large stretches of sky because their field of view is very small. But also they take usually hundreds of hours of observation. And so this is really like a taster of the immense deluge of data that awaits us in the next 10 years.

19:15And yes, a deluge of data is right. Like I'm sure hundreds of hard drives, thousands of hard drives are going to be filled up with information about the sky that's taken every, what, three or four days. Yes, and it will also form a detailed history of how the sky is changing. Now, a lot of astronomers are not going to actually look at any of the images, but they will receive nightly or daily alerts about, for example, supernova explosions happening in a distant galaxy. the observatory is expected to produce around 8 million of these alerts every night so you know a substantial amount of effort has gone into figuring out ways to kind of disseminate this information to astronomers in a way that doesn't overwhelm them and what of astronomers then david obviously this is an exciting time i'm sure for them with a new tool opening up ways that they can look at the universe.

20:11You've spoken to many of them. What have they told you? I think it's hard to convey how excited astronomers are because we're in the age of big data and astronomy has been changing because of that. But this observatory takes it to a whole other level. So no one has ever looked at the sky in such depth and with such exhaustive and repeated looks at the same patches of sky. So that means we'll be able to see fainter and fainter objects as the data accumulates, but also we'll be able to see how the sky changes over time. And so perhaps what astronomers are most excited about is that we'll be able to discover phenomena that no one even expected because we've never seen an evolving movie of the sky before.

20:53Well, I know that you'll be keeping an eye out for these evolving changes and nature will be reporting on them, I'm sure. But let's leave it there for the time being. Davide Castelvecchi, thank you so much for joining me today. Thank you very much. Nature's Davide Castelvecchi there. To read his article, look out for a link in the show notes. Finally on the show, it's time for the briefing chat. Now, every weekday, Nature publishes an email newsletter called The Nature Briefing, rounding up all the latest science news. We always have a read and we choose a couple of stories which have caught our eye to talk about.

21:21So, Ben, I'm going to go first this week. I'm very excited. I've got a story about a new kind of nitrogen. A new kind of nitrogen. So nitrogen, famously, what, N2? There's a load of it in the air around us, but this is not that. No. So this is a Chemistry World article about a Nature paper. And the new type of nitrogen is It's called hexanitrogen. Can you guess how many nitrogen atoms are in it? Okay, so hang on. Pentahexa, that's six, right? Ben, I can't believe it took you that long. Listeners can't see me on my fingers trying to work it out. Yeah, as you said, like dinitrogen, as it's called, is like the normal N2 that's in the air.

21:56It's super stable and it's also super strong. It's got like the two atoms with a triple bond between them, which makes it really stable and hard to break. But everyone's, for a long time, I've been very interested in making different kinds of nitrogens, what are called allotropes. And the most familiar ones are different allotropes of carbon. So basically, if you think there's carbon, then there's diamond, which is the sort of 3D cubic carbon. Then there's graphene, which is flat. Oh, and then there's Buckminster fullerene. Let's make weird balls of carbon. But they're all the same element. They're all pure carbon, but different forms.

22:35So that's what's called allotropes. So people have been really interested in making a new allotrope of nitrogen. And people have done it before. Previously, it's been only really, really high pressure that they've been able to make some sort of like weird multi-atom nitrogen. This one, not at high pressure, although it does have to be cryogenically frozen. It has to get really, really cold in order for the hexanitrogen to actually not break down. Well, there's always a kicker, isn't there, Shalmey? So what sort of temperature are we talking about? Yeah, so it's right down at liquid nitrogen temperatures.

23:06And basically, they have to make this stuff and then cool it really quickly. At room temperature, the molecule's lifetime is around 36 milliseconds. Wow. Whereas if you get it to cool, then you can get it to stick around. And you know what I was saying about the bonds? Dinitrogen, two nitrogen atoms, triple bond, super strong. One of the reasons it's really hard to make different forms of nitrogen is that it tends to break up into N2s. like that's a really stable thing for it to turn into. So part of the structure of this thing, and they've got little pictures of it, of the six atoms together.

23:39And one of the researchers, they're organic chemists, he described it as it's more like two times N3 than three times N2. Okay, right. That makes sense. So it's two little triplets together with a sort of weak point where they connect. It's only got a single bond where they connect and it's got like double bonds everywhere else. so it's not likely to then break down into N2 which is why it's like a little bit more stable than other things that people have been looking for working on trying to create Right, so is it like a string then rather than a ring or something like that? Yeah, from the pictures if you imagine two little triplets not in a line but sort of like slightly offset from each other and they've got sort of electron maps and all that sort of thing and yeah, so it's harder to break down but if you can get it to break down it has this huge amount of energy in So nitrogen in general, especially with these strong bonds that it makes, can store and release tons of energy.

24:34If you think about like TNT, like a lot of explosives are nitrogen based for that reason. So I get the feeling from reading this that the organic chemists who made hexanitrogen are mainly like, wow, we've made hexanitrogen. Like here's how we've done it. They want to go and make, I don't know what you call it, N10. They want to go and make like the next maybe slightly more stable form of nitrogen. But there are potentially practical applications. Energy storage, high density energy storage is kind of the one. So if you think about maybe using it for rocket fuel or something like that, because it would just create this huge burst of energy and the product that comes out would be nitrogen gas.

25:14So it's not putting loads of carbon dioxide into the air. So it could potentially be quite useful. Yeah, potentially, I suppose. But it seems like quite a difficult one to make. How do they go about it? Yes, it's really complex organic chemistry to make it. You know, lots in the paper, obviously, about like how they've done it. And some people at least are pretty impressed. So one of the researchers who was quoted, who's interviewed for this Chemistry World article, very effusive, said this work is spectacular and, in my opinion, is worthy of a Nobel Prize. Goodness. It's certainly very interesting and fun.

25:44And like I said, like N10 is the next allotrope of nitrogen that they have their eyes on. They've theorised what might be possible and that's the one that they think they can do. Well, that is cool research. Very cold research, I suppose. But let's move on to our second story this week in the briefing chat. And, well, it's a story that finally gives us an idea of what our extinct ancient relatives Denisovans might have looked like. And this is thanks to research published in the journals Cell and Science and a story I read about in Nature. I've definitely heard about this, and I was quite excited because I've definitely, over the years, heard a lot about Denisovans.

26:24And I get the idea that they've been quite mysterious. Absolutely right. Denisovans are, as I say, an extinct relative of ours, quite distinct from modern humans and Neanderthals. And you're right, you know, we don't really know much about them. And this research maybe puts to bed a decade and a half of speculation as to what Denisovans' appearance was. And it's been a mystery, really, since scientists identified them from unique DNA taken from a finger bone in a Siberian cave in 2010. So people have tried to use genetics and so forth to try and maybe reconstruct what a Denisovan might have looked like.

27:01But this research, it all comes down to a very particular skull, the cranium. So this is the upper portion of the skull lacking a jawbone. And this is one of the best preserved archaic human skulls. It was described in 2021, thought to be 146 ,000 years old. And there was a school of thought that this may have been a Denisovan, but evidence had been lacking. So this skull was originally described in 2021. Is this a new paper on it? Yeah, two new papers. And this skull, I mean, it's got quite a cool backstory. So a paleontologist in China obtained it from an unnamed man in 2018 who said that his grandfather had unearthed it in the early 1930s during the construction of a bridge and buried it in a well where it remained until his deathbed confession.

27:51and researchers obviously been looking at this but in this work in these two papers a team of researchers including one who worked on getting dna from the first denisovan finger bone they really wanted to see if they could get ancient dna from this skull to learn a bit more about it now they didn't they couldn't get any ancient dna directly from the bone but they did get some fragments of ancient proteins oh yeah 95 ancient proteins in fact and one of these the sequence of this protein matched one from the Denisovan finger bone and from other Denisovan fossils. But this protein was different from modern humans and Neanderthals, suggesting that this skull belongs to a Denisovan.

28:33And that must be such an exciting step up from only having like tiny bones like a finger to represent this entire ancient human lineage. Yes, and I think it opens up so many potential avenues. But there is more to this research, in fact. So, I mean, it turns out that this is the second time this year a Denisovan has been identified by ancient proteins. Like it really seems like it's a growth area, but there's even more research going on here. So the team found this evidence. There were a couple of proteins that were less conclusive matches, but they wanted to take things further. And further evidence of this being a Denisovan has come from a rather unexpected place.

29:10And that place is dental plaque. Oh, OK. So bits of old bacteria stuff that gets left behind between your teeth. Yeah, the very stuff. And dental plaque can be really, really tough. Anyone who's been to the dentist and have them kind of scrape it off will know that it can be a tough job. Now, we don't know what dental provisions were like 146 ,000 years ago, but it seems pretty good because this skull, not a lot of plaque. But the researchers managed to get a tiny, tiny, tiny little bit. And they were looking for host DNA within the microbial DNA. So this plaque is made by bacteria. So they were looking for ancient human DNA sort of trapped within it.

29:51And they found it. They found genetic sequences from maternal mitochondria that most closely relate to early Denisovans in Siberia. So now we have the ancient proteins and this extra ancient DNA from the dental plaque. I'm glad they found dental plaque, but I am also very impressed at the ancient humans' ability to keep their teeth clean without flossing. Like, I do want to know their secret. I mean, that's the real story, right? I mean, that's what we all want to know. That's what we should be focusing on here, dental hygiene. Oh, I mean, of course. But one thing this work does do is it gives us a sense of what Denisovans look like.

30:25Because that's been this big question, right? So we actually have this skull and researchers now say we have shown strong evidence that this is a Denisovan. What does it look like? Well, this skull, prominent brow ridge, a brain as large as modern humans and Neanderthals. And I guess we can sort of compare and contrast with Neanderthals and other extinct hominin. we know a bunch about them but very little about these mysterious Denisovans and as you've said now this gives a chance to really learn a bit more about them so it may be that there's other Denisovan fossils that have been discovered years ago in museums or wherever they may be and you can't get any dna or any proteins from them for whatever reason yeah now you can at least hold them up and say well these look kind of similar and from that we can maybe even find out more about where they lived and you know all these sorts of things that people are interested to know because i guess this skull didn't come with a load of associated context because it was discovered so long ago and then kept but if this then opens up the museum collections to like wait a sec we've got more yeah that's a yeah potentially a whole wealth of information yeah absolutely right and also it shows that this dental plaque method could be used to identify lineages of other ancient fossils as well so this getting stuff from plaque thing this is quite a new thing as well then So the researcher behind this says this is the first time that host DNA has been recovered from this plaque material from the Paleolithic era, which ended 12 ,000 years ago.

31:50So really, it turns out that there's all these new methods being made up to try and date stuff and work out where it is and all the rest of it. And I think we've covered a few of them. And I think Denisovan's is a story we will continue to cover because it is fascinating knowing that there was this species that no longer exists that really we don't know too much about. And I'm interested in further revelations on the dental hygiene of ancient humans. So I'll be keeping my eye on this slack method of paleontology. Well, absolutely. And let's leave it there for this week's briefing chat. Listeners, if you'd like to know more about those two stories, head over to the show notes for some links.

32:23And we'll also have a link on where you can sign up for the nature briefing to have even more stories like this delivered directly to your inbox. And also in the show notes, you'll find a link to our listener survey. And thank you so much to those of you who filled in already. If you've got a couple of minutes spare, head over there and tell us a bit more about who you are. If you're on Spotify, just scroll down and you can do it right there. And if you want to keep up to date with everything we're doing, why not give us a follow on social media? If you're on the social medias, we're at Nature Podcast.

32:53If you're there on your podcasting app, why not drop us a review? Give us some stars or a comment. We love those. And if you want to get in touch with us directly, we are podcast at nature.com. I'm Sharmini Bundel. And I'm Benjamin Thompson. Thanks for listening.

33:33Uncover new investment opportunities for you at creativeplanning.com slash new.

34:03Grainger for the ones who get it done.