AI linked to boom in biomedical papers, infrared contact lenses, and is Earth's core leaking?
20:11 Share

Did you know that foreign investors are quietly funding lawsuits in American courts through a practice called third-party litigation funding? Shadowy overseas funders are paying to sue American companies in our courts, and they don't pay a dime in U.S. taxes if there is an award or settlement. They profit tax-free from our legal system, while U.S. companies are tied up in court and American families pay the price to the tune of $5,000 a year.

But there is a solution. A new proposal before Congress would close this loophole and ensure these foreign investors pay taxes, just like the actual plaintiffs have to. It's a common sense move that discourages frivolous and abusive lawsuits and redirects resources back into American jobs, innovation, and growth. Only President Trump and congressional Republicans can deliver this win for America.

and hold these foreign investors accountable. Contact your lawmakers today and demand they take a stand to end foreign-funded litigation abuse.

NetSuite brings accounting, financial management, inventory, HR into one suite to help you know what's stuck, what it's costing you, and how to pivot fast. If your revenues are at least in the seven figures, download the free e-book, Navigating Global Trade, Three Insights for Leaders, at netsuite.com slash tech. ♪

Welcome back to The Nature Podcast. This week on the show, we're doing something a little bit different. We've got two brilliant hosts with me today as we talk about a couple of papers that have been highlighted in The Nature Briefing. So I am your host, Nick Petridge-Howe, and with me today is Sharmilee Bundell. Yep.

I'm here, ready to talk about some science. And Lizzie Gibney. Hello, good to be here. Well, it's great to have you both here as well. So who wants to go first today? We've got lots of exciting science. Sharmley's looking at me. Me, me, I'm looking keen, I'm looking keen. I'm really into this paper. Not that it's found a good thing, but I think it's a really interesting study. So I've been reading about this in Nature and it's an article that was published in PLOS Biology. And it's about...

Well, it's about maybe AI being used. Well, it's hard to prove, right? Has AI written a bunch of papers? Well, they're not saying like, yep, look for the telltale signs. This was definitely AI. But they are looking at some patterns. They've analysed over 300 papers that have used a specific data set. And they're looking at some patterns.

And they've come up with some reasons that they think it could well be that we are currently seeing basically an explosion of poor quality AI papers using this is one particular public data set. And this is always like the fear, isn't it, that AI is going to be used to like churn out a bunch of papers. I think, Lizzie, you've written about this previously.

But I guess it is hard to prove and maybe someone will be using AI just to help with their language or, you know, to help them sort of write more succinctly or something like that, which maybe is okay. And, you know, maybe some of the papers that they've looked at were made by people who looked at the same data set and sort of did the same kind of thing. But there's sort of a statistical problem here with the fact that you can use AI to produce a large number of analyses. Yeah.

So what this is about is the US National Health and Nutrition Examination Survey. So this is an open data set of health records. It looks at various things like diet and sleep and sort of lifestyle things and also then medical health conditions. And the authors described it as a sudden explosion in publication rates of what they describe again as extremely formulaic papers using this data. And they all seem to follow this template of there'll be a particular vet

variable like vitamin D levels or sleep quality. And then they'll associate that with a really complex disorder like heart disease or depression, which are things that have multiple causes and contributing factors. But they're looking for this association. And the problem with the statistics is if you've got this open data set,

You can run loads and loads of tests, theoretically, until you find one that has a positive association. So this is AI, so it's not being used to write the papers, not finding, you know, those telltale words that suggest that maybe someone's used AI to write something. It's actually maybe AI or some kind of automatic process was involved to actually trawl information

this data and kind of generate these formulaic papers on a much wider scale? Yeah, I mean, it could be writing it as well. But I think that, yeah, the problem is the data analysis. The problem is that this data set is openly available and kind of ready to plug your AI in and do your analysis on it.

it. What that means is that you can theoretically keep looking until you find something that's positive, that's positive association. So for example, one of the things that they found was a lot of these papers didn't use the full data set. They used like a subsection. So it might be like, we're going to look at this condition, but within these ages.

within these years or something like that. And there wasn't a particular reason given often in the papers for picking that. And so, for example, they looked at 14 papers which had links between a particular marker of blood inflammation and various conditions, including diabetes, hearing loss. Of those 14 papers, four of them used the complete data set. So why are the other ones sort of limiting their scope of what they're looking at? Well, it could be that people

has been sort of deliberately omitted because if you use the full data set, you don't get that result. Or it could be that, well, if you pick, let's say like you break up all your data into like five different year ranges, maybe you could get five papers out of that.

Or it could be that if you break up into five chunks, one of them is more likely to have a statistical positive result. Yeah, and that's the thing with the statistics, right? It's like it actually really matters whether you're going in and looking for a particular thing or not. If you think there's a link between vitamin D and sleep or whatever and you go in and look for that and you find a result, great. But if you're looking at every variable with every other variable and you're chunking out the data into the little bits that you want...

your findings mean absolutely nothing. You're probably going to find something there that seems to be correlated. Classic p-hacking. Exactly. There's a really nice analogy in here from one of the authors who says, imagine you're trying to pass an exam that has a particular pass rate and you add as many questions as you want, see which ones you got right and remove the ones you got wrong. That's basically what they're doing. And why are they doing this? Well, they've gotten...

A lot of, well, I say they, this is not, there's not, there's not a shadowy cabal here. But there have been, I shall say, a lot of publications. And the other thing, the other reason that these authors are sort of thinking like this might be AI linked is the rapid increase in the number of papers seems to be associated with this sort of availability of AI. So they found that kind of the publication rate started to increase markedly in 2022.

And in 2024, there were more than 2,200 association studies using this particular data set published. And so far this year, as of early May, it's already more than half that has already been published, again, using this particular data set. And

you've got papers that are being published in a range of journals Frontiers Media, El Sevier, Spring and Nature they've looked at. So an easy way to get lots of papers almost. Exactly I mean it's getting published and I should say one of the publishers that they looked at is Spring and Nature so it's important to note that Nature's news and podcast teams are

editorially independent of our publisher. And it's so difficult, isn't it? Because on one hand, this database is such an asset for researchers. You know, we have all this open data and there's been so many campaigns over the years to make data open so people can, it doesn't go to waste and everyone can use it and find whatever is there to be found. But then this is now exploiting that.

So what do you do? One of their suggestions was that public databases like this one should ask researchers to register their study plans before they'll give them access to the data.

So you can kind of audit it and stop what they describe as people wholesale mining these kind of data sets. Because when they're exploited like this, one of the authors says it drowns out any meaningful finding. Certainly, something needs to be done because all of these single association, you know, A is associated with B, and there's just sort of this proliferation of them. They're all kind of questionable, because these are complex things. And you're not

really getting very useful then information out of that and it's not the kind of thing where ah this is going to lead to a treatment because we found you know the smoking gun there probably isn't a smoking gun in all these cases yeah but it could just lead to you know misinformation or people wasting time building on these studies because if you don't see it in this big wide context you might just think oh yeah look there's a yeah a reasonable sounding association now i'm going to now base my next couple of months research on this and i'm

they might be completely wasting their time. Yeah. And I think this also raises the question of, is the problem here the proliferation of these kind of studies? Is there a problem with the fact that they're getting published? Where should the filter almost be in terms of judging the quality of papers and what information should be out there? No, as Lizzie says, this is something that's important to not mislead

the field but you know when you were talking about smoking guns I was thinking about smoke seeing clearly this is a terrible thing but I know that it is a segue it is a segue segue alert but I know that Lizzie has been looking at something a little bit to do with vision with some contact lenses that's right bringing a little bit of sci-fi to the pod almost so these are contact lenses that let you see in the infrared

So humans can't normally see in the infrared unless you haven't noticed that. Some animals kind of can, but like not very well. They more like just sense it. So this is, you know, in one way, superhuman vision. It's a bit like wearing night vision goggles because they are also able to see in the infrared. But these are contact lenses. And there's different kinds of infrared, isn't there? Because some animals...

like I think snakes for example are very much sensing heat so you can get like thermal imagery which will let you like people glowing that's slightly different so if you think of because we you know most of our information about this probably comes from like films about the army yeah

So you have night vision goggles. You have the ones where you just see in all that kind of green monochrome. And that's just seeing at night. And that can be very useful. And then you have the thermal imaging as well, which is, you know, if you're doing a mission where you need to, there's a hostage or something, you know, and you can see their heat, their body heat. And that is the very far end of the infrared. So that is also useful.

That's harder to see in this way. So what we're talking about is the near infrared. So more the just normal night vision. And how do you make contact lenses that give you super night vision? You use some nanoparticles. So these are, as the name would suggest, very, very small, about 45 nanometers across. And they are made with some rare earth metals inside. And they take the incoming infrared light.

which is a longer wavelength and they absorb it and they spit out wavelengths of light that we can see. So shades of green and blue and red they can do. And they can actually layer them as well. So they can kind of input three different infrared wavelengths and output three different visible light wavelengths. So it's just a little conversion process.

But as you can imagine, there are a few little caveats. So I mentioned night vision goggles because it is a great comparison, right? But there is quite a big difference. Night vision goggles are big and clunky and bulky for a reason because they have power sources in them. They have batteries in them. And part of that is because they amplify the light coming in.

The contact lenses do this completely passively, so that's great. There's no batteries involved, but they just convert what they get coming in, they convert it into the visible. The problem is there isn't always a huge amount of infrared light. So what they're able to see or the participants in the study were able to see were just very bright sources. Well, I say very bright LED kind of sources. So if you were looking in a dark room at night, you wouldn't

see anything but if there was a little infrared led flashing which you wouldn't see normally without contact lenses you would see with the lenses in and you said participants they tried this in people they did they tried it in mice and in people so with mice there are two boxes

And one was lit up in the infrared and one that was dark. And ones with lenses always ran to the one that was dark. And the ones without lenses, they both looked dark to them. So they ran equally into either. And in people, when wearing the lenses, the people were able to detect enough of this LED light and the direction of it to kind of get Morse code signals.

Yeah. So they were. Exactly. Yes. Yes. Oh, the other thing is that they also do make things a little bit blurry because the nanoparticles are, they worked really hard, the team, to make them as seamless as possible. So to not change, you know, the transparency of the lenses so you can still see visible light. Absolutely fine. Yeah.

but they still scatter the light when it comes in a little bit. So the vision is not very clear when you're wearing them, but they found a way to get around that as well. They can kind of put the lenses in a more traditional kind of glasses format and then they can redirect the light. And then the participants were able to read letters that otherwise wouldn't have been visible. And you said this is like only really working with very bright lights. What might it be used for? Because you can't do the sort of thing that you said where you like see people through walls and like save the hostage or whatever. Yeah.

That's right. So, well, as you mentioned vision at the beginning and smoke, if you're in a search and rescue situation, it could be very dusty in the air. Infrared travels through dust much better than visible light does. And for that same reason, actually, you can see with these lenses on even with your eyes closed because infrared does go through your eyelids where invisible light does not to the same extent. So it's a similar thing as you get much better penetration with infrared.

So that's one possibility. You can also have, again, it very quickly starts to be like spy stuff, but you can have like anti-counterfeit marks on, I don't know, cash or products or whatever that you can read, you know, you can see an emitting ink in infrared and you can see them with the contacts in that you wouldn't be able to without ink.

And actually, one person I spoke to for this story, which I should have said is my story. I'm not sure if I said that. I wrote about this in Nature. And it was published in Cell. And one of the sources that I spoke to suggested that it could also be used in surgery. So you have, again, a kind of an emitter that you can put into a cancerous lesion and then it will emit in the infrared. And a surgeon could be wearing contact lenses and that might help them to very easily direct, take out the right bit of

so there are quite a few uses they they probably will need to be improved quite a bit but it's a it's an incredible proof of principle that it's even possible really and i should say as well that in addition to writing a news story you also did a little short video which will be on the nature youtube channel i'll link to it in the show notes so you're not wearing the lenses but you are talking about them and you can see the lenses in the video exactly exactly and yeah they are i wish i could try the lenses that would be super cool and apparently they're

only about $200 to make. So I was like, that's not that bad. Although I don't know if they're the contact lenses that are quite expensive. Maybe then it is expensive. I feel like it's the kind of thing that small children would like for their secret messages. No.

Not the best use, but that's what I'm thinking. Yes, with the invisible ink as well. Yeah, exactly. A lot of potential. A lot of potential, not only for search and rescue, but also for children's parties. Very cool story, and I will link to it in the show notes. And then finally on this episode, we've got my story this week, and it asks the question...

Is Earth's core leaking? That sounds worrying, doesn't it? That's exactly what I thought. Wait a sec. No, Earth's core is leaking. Isn't that what volcanoes are? Oh, that's a very good point. No, that's the mantle. Yes, I was going to say that's the mantle. The Earth's core you would expect to be very well wrapped up and very well contained within the centre of the planet. I'm going to say no, no it's not.

Is that right? The answer is maybe. This article that I was reading in Nature, about a Nature paper in fact, was written by Davide Castelvecchi, who often comes on the show, and it shows the strongest evidence yet that the Earth's core would be leaking. Wow.

leaking how do you mean leaking and what so what's different about what's in the core compared to what's the normal stuff that leaks out the difference with the earth's core is it was formed before everything else so it being at the center of the planet means it was formed early in earth's formation so the solar system itself was formed about 4.5 billion years ago and then the core was formed a little bit later than that and then the mantle which is a bit between the core and the crust that was formed after that and then finally the crust when it

all cooled at the end. So the difference would be that there would be different isotopes of atoms in the core that would be more similar to what was there in the early solar system than that would be in the mantle and the

the crust. So to answer your question, Sharmilee, it's not worrying. If it is indeed leaking, it's been leaking for a very long time and so far things are fine. And the way in which it would be leaking would be, as you say, through volcanoes, but through a very specific one that seems to have run very deep. So this was a study looking at volcanic rocks in Hawaii, where it's already known that

that there is a magma stream coming from the very deep mantle and now with this evidence they think that it's also coming from the core itself. So there must be like a crack somewhere in the core that's leaking out into the mantle and then the mantle is leaking out through the crust. It must be mixing with the other stuff. It's not like there's a straw going straight down to the middle and it's all...

One direct pipeline. Yeah. So at the boundary, there's some sort of leakage that is between the two. I don't know if it's one specific place or it's all across the core. But according to this study, at least, there is good evidence that there is a leak from the core. So in this analysis, they were looking at a particular element called ruthenium. And this is an element that is very rare, but is thought to be found in great quantities in minerals.

the core or at least greater quantities it's still quite rare and so what they did is they developed a way to analyse the very tiny amounts of it you might get in rocks that they collected from Hawaii because in the past analyses they've looked at they've not been sensitive enough to pick up the very tiny differences that you'd expect from

And then they looked at the different isotopes and found that actually the isotopes they were finding in these rocks in Hawaii seemed to be what you would expect to find in the core, which is good evidence for there being a bit of core leakage. And this would have been some time ago to be long enough for those rocks to then form from...

eruptions. Yeah, the islands of Hawaii were formed a long time ago, not that long ago in Earth's history, but a long time ago to humans like us. And so it's come out and through volcanoes and all the rest of it and formed these islands. So it's been quite a while ago. And yeah, it would have been material coming from about 3,000 kilometers below the surface, so quite a long way. And can we learn anything about the core?

from this because obviously it's by definition pretty inaccessible but you know I'd be fascinated to know what's what's going on in the core like does it just confirm you know that it's made of the things we already knew? Well there's been this idea that maybe the core is leaking for a little while but the evidence hasn't been compelling enough so past studies have looked at tungsten as well which is supposed to have certain kinds of isotopes that come from the core but

This basically, if it is true, would sort of upend our textbook understanding, because the understanding is that the core is a separate thing, then there's the mantle, and at these points nothing shall cross. But actually, maybe there is a bit of crossing going on, and what that means, yet to be determined, and...

again, this is the strongest possible evidence yet. It's not for sure this is happening. So more work to understand if this is actually happening and then we can start throwing out the textbooks and sort of thinking about what this all means. I love the idea of some high school kids listening to this and complaining to their teacher that the textbooks are not right. It's mixing. Haven't they, Castle Vecchi told us. I'm just always like throwing out textbooks.

It's like, you know you succeeded in science where you're like, throw away the textbooks. They're out of date. I've done new science. That's a really interesting one. So I'm glad we're safe from the mixing, but could tell us some interesting geological facts about the formation and centre of the Earth.

And no doubt there will be a link to that story in the show notes. There will be links to all of the stories and the video that we discussed in this week's show notes. And also we'll put a link of where you can sign up to the briefing to get more stories like these directly to your inbox. But for now, I think that's all we've got. Wonderful. Well, don't forget, you can also follow The Nature Podcast on Instagram.

X and blue sky and email. We're at nature podcast on social media or podcast at nature.com. If you want to email us. And if you enjoy listening to us talk about science, then you can leave us a review wherever you get your podcast. I appreciate five star reviews, but you know, whatever you feel about the podcast is totally fine. But for now, thank you so much for listening. I'm Nick Petrichow. I'm Sharmini Bandal. And I'm Lizzie Gibney.

Did you know that foreign investors are quietly funding lawsuits in American courts through a practice called third-party litigation funding? Shadowy overseas funders are paying to sue American companies in our courts, and they don't pay a dime in U.S. taxes if there is an award or settlement. They profit tax-free from our legal system, while U.S. companies are tied up in court and American families pay the price to the tune of $5,000 a year.

But there is a solution. A new proposal before Congress would close this loophole and ensure these foreign investors pay taxes, just like the actual plaintiffs have to. It's a common sense move that discourages frivolous and abusive lawsuits and redirects resources back into American jobs, innovation, and growth. Only President Trump and congressional Republicans can deliver this win for America.

and hold these foreign investors accountable. Contact your lawmakers today and demand they take a stand to end foreign-funded litigation abuse. Hi, guys. It's Hannah from Giggly Squad. With summer on the corner, I wanted to tell you guys how I'm staying comfy and safe.

stylish. Lululemon is my secret weapon. There are plenty of copycats out there, but nothing compares to the Lululemon fabrics and fit. I've literally had my pair of Lululemon leggings since college and I'm out of college. I know I don't look it, but I am. The quality is next level. I especially love the Lululemon Align collection. It's made with this weightless, buttery, soft Nulu fabric that feels like next to nothing. It's so soft. It's

Whether you're in Align pants, shorts, a bra, tank, skirt, a dress, you get nonstop flexibility in every direction so you can stretch the summer limits. Align even wicks sweat. And as a sweaty girl, I love this. You know it's going to be my best friend when I play tennis this summer. Shop the Align collection online at lululemon.com or your nearest Lululemon store.