Little Deming Lake doesn’t get much notice from visitors to Itasca State Park^[1] in Minnesota. There’s better boating on nearby Lake Itasca, the headwaters of the Mississippi River. My colleagues and I need to maneuver hundreds of pounds of equipment down a hidden path made narrow by late-summer poison ivy to launch our rowboats.

But modest Deming Lake offers more than meets the eye for me, a geochemist^[2] interested in how oxygen built up in the atmosphere 2.4 billion years ago. The absence of oxygen in the deep layers of Deming Lake is something this small body of water has in common with early Earth’s oceans.

On each of our several expeditions here each year, we row our boats out into the deepest part of the lake – over 60 feet (18 meters), despite the lake’s surface area being only 13 acres. We drop an anchor and connect our boats in a flotilla, readying ourselves for the work ahead.

Deming Lake is meromictic^[4], a term from Greek that means only partially mixing. In most lakes, at least once a year, the water at the top sinks while the water at the bottom rises because of wind and seasonal temperature changes that affect water’s density. But the deepest waters of Deming Lake never reach the surface^[5]. This prevents oxygen in its top layer of water from ever mixing into its deep layer.

Less than 1% of lakes are meromictic, and most that are have dense, salty bottom waters. Deming Lake’s deep waters are not very salty, but of the salts in its bottom waters, iron is one of the most abundant^[6]. This makes Deming Lake one of the rarest types of meromictic lakes^[7].

The lake surface is calm, and the still air is glorious on this cool, cloudless August morning. We lower a 2-foot-long water pump zip-tied to a cable attached to four sensors. The sensors measure the temperature, amount of oxygen, pH and amount of chlorophyll in the water at each layer we encounter. We pump water from the most intriguing layers up to the boat and fill a myriad of bottles and tubes, each destined for a different chemical or biological analysis.

My colleagues and I have homed in on Deming Lake to explore questions about how microbial life adapted to and changed the environmental conditions on early Earth. Our planet was inhabited only by microbes^[9] for most of its history. The atmosphere and the oceans’ depths didn’t have much oxygen, but they did have a lot of iron, just like Deming Lake does. By investigating what Deming Lake’s microbes are doing, we can better understand how billions of years ago they helped to transform the Earth’s atmosphere and oceans into what they’re like now.

Layer by layer, into the lake

Two and a half billion years ago, ocean waters had enough iron to form today’s globally distributed rusty iron deposits called^[10] banded iron formations^[11] that supply iron for the modern global steel industry. Nowadays, oceans have only trace amounts of iron^[12] but abundant oxygen. In most waters, iron and oxygen are antithetical. Rapid chemical and biological reactions between iron and oxygen^[13] mean you can’t have much of one while the other is present.

The rise of oxygen in the early atmosphere and ocean was due to cyanobacteria^[14]. These single-celled organisms emerged at least 2.5 billion years ago^[15]. But it took roughly 2 billion years for the oxygen they produce via photosynthesis to build up to levels that allowed for the first animals^[16] to appear on Earth.

At Deming Lake, my colleagues and I pay special attention to the water layer where the chlorophyll readings jump. Chlorophyll is the pigment^[18] that makes plants green. It harnesses sunlight energy to turn water and carbon dioxide into oxygen and sugars. Nearly 20 feet (6 meters) below Deming’s surface, the chlorophyll is in cyanobacteria and photosynthetic algae, not plants.

But the curious thing about this layer is that we don’t detect oxygen, despite the abundance of these oxygen-producing organisms. This is the depth where iron concentrations start to climb to the high levels present at the lake’s bottom.

This high-chlorophyll, high-iron and low-oxygen layer is of special interest to us because it might help us understand where cyanobacteria lived in the ancient ocean, how well they were growing and how much oxygen they produced.

We suspect the reason cyanobacteria gather at this depth in Deming Lake is that there is more iron there than at the top of the lake. Just like humans need iron for red blood cells^[19], cyanobacteria need lots of iron to help catalyze the reactions of photosynthesis.

A likely reason we can’t measure any oxygen in this layer is that in addition to cyanobacteria, there are a lot of other bacteria here. After a good long life of a few days, the cyanobacteria die, and the other bacteria feed on their remains. These bacteria rapidly use up any oxygen produced by still photosynthesizing cyanobacteria the way a fire does as it burns through wood.

We know there are lots of bacteria here based on how cloudy the water is, and we see them when we inspect a drop of this water under a microscope. But we need another way to measure photosynthesis besides measuring oxygen levels.

Long-running lakeside laboratory

The other important function of photosynthesis is converting carbon dioxide into sugars, which eventually are used to make more cells. We need a way to track whether new sugars are being made, and if they are, whether it’s by photosynthetic cyanobacteria. So we fill glass bottles with samples of water from this lake layer and seal them tight with rubber stoppers.

We drive the 3 miles back to the Itasca Biological Station and Laboratories^[20] where we will set up our experiments. The station opened in 1909 and is home base for us this week, providing comfy cabins, warm meals and this laboratory space.

In the lab, we inject our glass bottle with carbon dioxide that carries an isotopic tracer^[21]. If cyanobacteria grow, their cells will incorporate this isotopic marker.

We had a little help to formulate our questions and experiments. University of Minnesota students attending summer field courses collected decades worth of data in Itasca State Park. A diligent university librarian digitized thousands of those students’ final papers^[22].

My students and I pored over the papers concerning Deming Lake, many of which tried to determine whether the cyanobacteria in the chlorophyll-rich layer are doing photosynthesis. While most indicated yes, those students were measuring only oxygen and got ambiguous results. Our use of the isotopic tracer is trickier to implement but will give clearer results.

That afternoon, we’re back on the lake. We toss an anchor; attached to its rope is a clear plastic bag holding the sealed bottles of lake water now amended with the isotopic tracer. They’ll spend the night in the chlorophyll-rich layer, and we’ll retrieve them after 24 hours. Any longer than that and the isotopic label might end up in the bacteria that eat the dying cyanobacteria instead of the cyanobacteria themselves. We tie off the rope to a floating buoy and head back to the station’s dining hall for our evening meal.

Iron, chlorophyll, oxygen

The next morning, as we wait for the bottles to finish their incubation, we collect water from the different layers of the lake and add some chemicals that kill the cells but preserve their bodies. We’ll look at these samples under the microscope to figure out how many cyanobacteria are in the water, and we’ll measure how much iron is inside the cyanobacteria.

That’s easier said than done, because we have to first separate all the “needles” (cyanobacteria) from the “hay” (other cells) and then clean any iron off the outside of the cyanobacteria. Back at Iowa State University, we’ll shoot the individual cells one by one into a flame that incinerates them, which liberates all the iron they contain so we can measure it.

Our scientific hunch, or hypothesis^[25], is that the cyanobacteria that live in the chlorophyll- and iron-rich layer will contain more iron than cyanobacteria that live in the top lake layer. If they do, it will help us establish that greater access to iron is a motive for living in that deeper and dimmer layer.

These experiments won’t tell the whole story of why it took so long for Earth to build up oxygen, but they will help us to understand a piece of it – where oxygen might have been produced and why, and what happened to oxygen in that environment.

Deming Lake is quickly becoming its own attraction for those with a curiosity about what goes on beneath its tranquil surface – and what that might be able to tell us about how new forms of life took hold long ago on Earth.

Uncommon Courses^[1] is an occasional series from The Conversation U.S. highlighting unconventional approaches to teaching. Title of course:Diversity of Biological Sex CharacteristicsWhat prompted the idea for the course?Most people view biological sex, or the physical features related to reproduction, as simple and binary^[2] – either male or female. Even those who recognize that gender – referring to cultural norms around biological sex, or a person’s internal feeling of being masculine, feminine or both – can be complex and nuanced don’t see biological sex in the same way. Many also regard variability in sex and gender as exclusive to people – not found in nonhuman animals.I am a behavioral neurobiologist^[3] who has been teaching human physiology since 1998. Over the past several years, I have focused my reading and writing^[4] on the biology of sex^[5]. It struck me that many of my students had misguided assumptions about sex characteristics, including that all people are physically either 100% male or 100% female. A course on biological sexual diversity^[6] in both nonhuman animals and people could challenge these assumptions.What does the course explore?First, we examine why sexual reproduction evolved^[7] in any species. This question is still hotly debated among biologists^[8] because sex is inefficient. It requires time and energy to find a suitable mate and unite your sex cells, plus it allows you to pass on only half your genes to your offspring.In comparison, asexual reproduction^[9] – essentially cloning yourself – is much more efficient. You don’t have to find a mate, and everyone can produce offspring themselves because there are no males. In biology, “male” refers to an individual that makes small sex cells like sperm, and “female” refers to an individual that makes large sex cells like eggs.Next, we explore nonhuman sexual diversity^[10], including fungi that have thousands of sexes^[11] and aphids that reproduce asexually most of the year^[12] but sexually once each fall. Among many others, we also learn about fish that are male or female at different times of their lives^[13]; intersex crayfish^[14]; and female spotted hyenas^[15] that have a penis. We then transition from nonhuman animals to people, via the brain. We learn about a few small^[16] brain structures in vertebrates^[17] that likely have reproductive functions and are differently sized in females versus males on average. We also learn that most people have some brain structures that are more typically male, others that are more typically female and still others that are intermediate – in other words, most people are mosaics^[18] of female-typical and male-typical brain sex characteristics.Finally, we focus on the biological sex characteristics of intersex people^[19]. The chromosomes and reproductive organs^[20] of intersex people have some typically female and some typically male characteristics^[21] or are intermediate between them.Students then build on their knowledge of the diversity of biological sex characteristics to discuss whether intersex infants should have surgery to “correct” their genitals^[22], as well as who should be allowed^[23] to compete in^[24] girls and women’s athletics^[25].Why is this course relevant now?Perhaps more than ever, there is a debate about how to treat people who do not fit neatly into a female or a male box. Many assume that biological sex is binary and regard transgender and nonbinary people as mistaken or confused. In addition, for many decades, intersex infants^[26] have undergone surgical procedures to make them appear more typically male or female. Even those who support transgender, nonbinary and intersex people often assume that biological sex is binary. But this assumption is not anchored in evidence.What will the course prepare students to do?Students often say that before they took this course, they had no idea biological sex characteristics could be so diverse, despite having taken several biology courses. An improved awareness of the complexity of biological sex may help shape the research and teaching of future biologists. This will help them design experiments that take account of the diversity of their subjects and be more inclusive in their teaching^[27]. It may also help all students ask better questions and make better judgments about social and political issues related to sex and gender.

Curious Kids^[1] is a series for children of all ages. If you have a question you’d like an expert to answer, send it to This email address is being protected from spambots. You need JavaScript enabled to view it.^[2]. How many hours does the average American spend on devices each year? – Maxwell P., age 10 Think about your favorite devices – your smartphone, laptop, tablet, computer or console – the things you use to play cool games, watch hilarious videos and connect and chat with friends. Many young people spend a lot of free time looking at them. Turns out that teens spend an average of 8½ hours on screens per day^[3], and tweens – that’s ages 8 to 12 – are not far behind, at 5½ hours daily.Keep in mind those numbers are for only social media, gaming and texting. They do not include the time that kids used screens for schoolwork or homework^[4].What’s more, much of the time taken up by social media and texting is apparently not even enjoyable, much less productive. A 2017 study of teens ages 13 to 18 suggests they spend most of those hours on the phone in their bedroom, alone and distressed^[5]. These lonely feelings correlate with the rise in the use of digital media^[6]. In 2022, 95% of teens had smartphones^[7] compared with only 23% in 2011. And 46% of today’s teens say they use the internet almost constantly, compared with 24% of teenagers who said the same in 2014 and 2015. Our team of psychiatrists^[8] who treat young people with digital addiction have many patients who spend over 40 hours per week on screens – and some, up to 80 hours. Think about it: If you spend “just” an average of 50 hours per week on devices from ages 13 to 18 – the total time you will spend on screens equates to more than 12 years of school! Find the right balanceAll this is not to say that everything about devices is bad. In this digital age, people embark on exciting journeys through the screens of their devices. Sometimes, screens are the windows to a magical adventure. But too much screen time can lead to problems. As human beings, we function best when we’re in a state of balance. That happens when we eat well, exercise regularly and get enough sleep.But spending too much time using digital devices can cause changes in the way you think and behave. Many teens and tweens developed the “fear of missing out” – known as FOMO^[9]. And one study shows some people develop nomophobia^[10], which is the fear of being without your phone, or feeling anxious when you can’t use it. Moreover, digital addiction in high school may predict serious depression, anxiety and sleep disruption^[11] in college. Rates of depression and anxiety are skyrocketing among college students^[12]. The fear of missing out is pervasive, resulting in sleep disruption; too many college students sleep with smartphones turned on^[13] and near their bed – and wake up to respond to texts and notifications during the night. Sleep disruption itself is a core symptom of both depression and anxiety^[14]. How to avoid device addictionA 2016 poll indicated that half of teens felt they were addicted to their mobile devices^[15]. Getting hooked on screens means missing out on healthy activities. To achieve a better balance, some experts recommend the following^[16]: Turn off all screens during family meals and outings. Don’t complain when your parents use parental controls. And turn off all the screens in your bedroom 30 to 60 minutes before bedtime – this step will improve sleep. You may be a “screen addict^[17]” if you: Feel uneasy or grumpy when you can’t use your device. Don’t take breaks while spending hours on your device. Ignore other fun activities you enjoy, like going outside or reading a book. Have trouble sleeping, or falling asleep, because your screen time is too close to bedtime. Experience eye, lower back and neck strain. Struggle with weight gain or obesity because you’re inactive. Have difficulty with real-life, face-to-face social interactions. If you notice these signs, do not dismiss them. But also realize you’re not alone and help is out there. You can find balance again. A healthy approachExercise – riding a bike, playing sports, lifting weights or going for a jog or walk – keeps your brain healthy^[18] and protects it against depression and anxiety, as well as limiting your screen time.Another way to be happier and healthier is to spend time with people – face to face, not via a screen. Seeing people live and in person is the best way to bond with others^[19], and it may be even better for life span than exercise. Creative hobbies are good, too. Cooking, playing an instrument, dancing, any arts and crafts, and thousands of other fun things make people happier and more creative. What’s more, hobbies make you well rounded and more attractive to others^[20] – which leads to more face-to-face interactions. It’s also critical for parents to practice healthy screen habits. But about one-third of adults say they use screens “constantly^[21].” This is not exactly a great example for kids; when adults take responsibility to minimize their own screen time, the whole family gets better.Our research team used magnetic resonance imaging, also known as MRI^[22], to scan the brains of teens who had digital addiction. We found impairment in the brain’s decision-making, processing and reward centers. But after a digital fast – meaning the addicted teens unplugged for two weeks – those brain abnormalities reversed, and the damage was undone. Our findings also showed that kids with a desire to overcome digital addiction did better with a digital fast than those who were less willing or who denied their addiction. A balanced lifestyle in the digital age is all about finding joy in screenless activities – being active, connecting with others and exploring your offline interests.Hello, curious kids! Do you have a question you’d like an expert to answer? Ask an adult to send your question to This email address is being protected from spambots. You need JavaScript enabled to view it.^[23]. Please tell us your name, age and the city where you live.And since curiosity has no age limit – adults, let us know what you’re wondering, too. We won’t be able to answer every question, but we will do our best.

Google, Microsoft and others boast that generative artificial intelligence tools like ChatGPT will make searching the internet^[1] better than ever for users^[2]. For example, rather than having to wade through a sea of URLs, users will be able to just get an answer combed from the entire internet.

There are also some concerns with the rise of AI-fueled search engines^[3], such as the opacity over where information comes from, the potential for “hallucinated” answers and copyright issues.

But one other consequence is that I believe it may destroy the US$68 billion search engine optimization^[4] industry that companies like Google helped create.

For the past 25 years or so, websites, news outlets, blogs and many others with a URL that wanted to get attention have used search engine optimization^[5], or SEO, to “convince” search engines to share their content as high as possible in the results they provide to readers. This has helped drive traffic to their sites and has also spawned an industry of consultants and marketers who advise on how best to do that.

As an associate professor of information and operations management^[6], I study the economics of e-commerce. I believe the growing use of generative AI will likely make all of that obsolete.

How online search works

Someone seeking information online opens her browser, goes to a search engine and types in the relevant keywords. The search engine displays the results, and the user browses through the links displayed in the result listings until she finds the relevant information.

To attract the user’s attentions, online content providers use various search engine marketing strategies, such as search engine optimization^[7], paid placements^[8] and banner displays^[9].

For instance, a news website might hire a consultant to help it highlight key words in headlines and in metadata so that Google and Bing elevate its content when a user searches for the latest information on a flood or political crisis.

How generative AI changes search process

But this all depends on search engines luring tens of millions of users to their websites. And so to earn users’ loyalty and web traffic, search engines must continuously work on their algorithms to improve the quality of their search results.

That’s why, even if it could hurt a part of their revenue stream, search engines have been quick to experiment with generative AI^[11] to improve search results. And this could fundamentally change the online search ecosystem.

All the biggest search engines have already adopted or are experimenting with this approach. Examples include Google’s Bard^[12], Microsoft’s Bing AI^[13], Baidu’s ERNIE^[14] and DuckDuckGo’s DuckAssist^[15].

Rather than getting a list of links, both organic and paid, based on whatever keywords or questions a user types in, generative AI will instead simply give you a text result^[16] in the form of an answer. Say you’re planning a trip to Destin, Florida, and type the prompt “Create a three-day itinerary for a visitor” there. Instead of a bunch of links to Yelp and blog postings that require lots of clicking and reading, typing that into Bing AI will result in a detailed three-day itinerary.

Over time, as the quality of AI-generated answers improve, users will have less incentive to browse through search result listings. They can save time and effort by reading the AI-generated response to their query.

In other words, it would allow you to bypass all those paid links and costly efforts by websites to improve their SEO scores, rendering them useless.

When users start ignoring the sponsored and editorial result listings, this will have an adverse impact on the revenues of SEO consultants, search engine marketers^[17] consultants and, ultimately, the bottom line of search engines themselves.

The financial impact

This financial impact cannot be ignored.

For example, the SEO industry generated $68.1 billion globally in 2022^[18]. It had been expected to reach $129.6 billion by 2030, but these projections were made before the emergence of generative AI put the industry at risk of obsolescence.

As for search engines^[19], monetizing online search services is a major source of their revenue^[20]. They get a cut of the money that websites spend on improving their online visibility through paid placements, ads, affiliate marketing and the like, collectively known as search engine marketing. For example, approximately 58% of Google’s 2022 revenues^[21] – or almost $162.5 billion – came from Google Ads, which provides some of these services.

Search engines run by massive companies with many revenue streams, like Google and Microsoft, will likely find ways to offset the losses by coming up with strategies to make money off generative AI answers. But the SEO marketers and consultants who depend on search engines – mostly small- and medium-sized companies^[22] – will no longer be needed as they are today, and so the industry is unlikely to survive much longer.

A not-too-distant future

But don’t expect the SEO industry to fade away immediately. Generative AI search engines are still in their infancy and must address certain challenges before they’ll dominate search.

For one thing, most of these initiatives are still experimental^[23] and often available only to certain users. And for another, generative AI has been notorious for providing incorrect^[24], plagiarized^[25] or simply made-up answers^[26].

That means it’s unlikely at the moment to gain the trust or loyalty of many users.

Given these challenges, it is not surprising that generative AI has yet to transform online search^[27]. However, given the resources available to researchers working on generative AI models, it is safe to assume that eventually these models will become better at their task, leading to the death of the SEO industry.

The Nobel Prize in chemistry for 2023 goes to three scientists “for the discovery and synthesis of quantum dots^[1].” The Conversation Weekly podcast^[2] caught up with one of this trio, physical chemist Louis Brus^[3], who did foundational work figuring out that the properties of these nanoparticles depend on their size. Brus’ phone was off when the Nobel reps called to inform him of the good news, but now plenty of people have gotten through with congratulations and advice. Below are edited excerpts from the podcast.

When you were working at Bell Labs in the 1980s and discovered quantum dots, it was something of an accident. You were studying solutions of semiconductor particles. And when you aimed lasers at these solutions, called colloids, you noticed that the colors they emitted were not constant.

On the first day we made the colloid, sometimes the spectrum was different. Second and third day, it was normal. There certainly was a surprise when I first saw this change in the spectrum. And so, I began to try to figure out what the heck was going on with that.

I noticed that the property of the particle itself began to change at a very small size.

What you’d found was a quantum dot: a type of nanoparticle that absorbs light and emits it at another wavelength. Crucially, the color of these particles changes depending on the actual size of the particle. How do you even see a quantum dot crystal, since one is just a few hundred thousandths the width of a human hair?

Well, you can’t see them with an optical microscope because they’re smaller than the wavelength of light. There are ways to see them too, using other types of specialist microscopes, such as an electron microscope. And a common way of demonstrating them is to line up a row of brightly colored glass flasks each with a solution of different sized quantum dots inside it.

One of your fellow laureates, Alexei Ekimov^[6], was a Russian scientist, and he’d actually observed quantum dots in colored glass, but you weren’t aware of his findings at the time?

Yes, that’s right. The Cold War was going on at that time, and he published in the Russian literature, in Russian. And he wasn’t allowed to travel to the West to talk about his work.

I asked around among all the physicists, was there any work on small particles? I was trying to make a model of the quantum size effects. And they told me no, nobody’s really working on this. Nobody had seen his articles, basically.

I was part of the U.S. chemistry community, doing synthetic chemistry in the laboratory. He was in the glass industry in the Soviet Union, working on industrial technology.

When I eventually found his articles in the technological literature, I wrote a letter to the Soviet Union, with my papers, just to say hello to Ekimov and his colleagues. When the letter came, the KGB came to talk to the Russian scientists, trying to figure out why they had any contact with anybody in the West. But in fact they had never talked to me or anyone in the West when my letter arrived in the mail.

Have you met him since?

Yes, they were able to come out of the Soviet Union during Glasnost, this would be the late 1980s. There’s Ekimov, and then there is his theoretical collaborator Sasha Efros^[7], who now works at the U.S. Naval Research Lab^[8]. I met them as soon as they came to the U.S.

---

Listen to the interview with Louis Brus on The Conversation Weekly podcast. Each week, academic experts tell us about the fascinating discoveries they’re making to understand the world and the big questions they’re still trying to answer.

---

One of the issues with quantum dots, when you first observed them, was how to actually produce them and keep them stable. Then, in the 1990s, your fellow laureate, Moungi Bawendi^[9], figured this out. What do you think is the most striking thing that you’ve seen quantum dots used in so far?

Usually when a new material is invented, it takes a long time to figure out what it’s really good for. Research scientists, they have ideas, you might use it for this, you might use it for that. But then, if you talk to people in the actual industry, who deal every day with manufacturing problems, these ideas are often not very good.

But the knowledge that we gained, the scientific principles, could be used to help to design new devices.

As far as first applications, people began to try to use them in biological imaging. Biochemists attach quantum dots to other molecules to help map cells and organs. They’ve even been used to detect tumors, and to help guide surgeons during operations.

And as scientists kept working to synthesize quantum dots, the quality of the particles kept improving. They were emitting pure colors, rather than distributions of light – like maybe red with a little bit of green, or maybe red with some pink. When you got a better particle, it would be just pure red, for instance.

So then people made the connection to the display industry – computer displays and television displays. In this application, you want to convert electricity into three colors: red, green and blue. You can make up any kind of image, starting with just those three colors in different proportions.

It takes a lot of courage. You have to invest a lot of money to develop the technology, and maybe at the end of it, it’s not good enough, and it will not replace what you already have. And there’s a lot of credit due to the Samsung Corporation in Japan. Hundreds of billions of dollars were invested in the technology of these particles to get them to the point where they could begin to manufacture displays and flat-panel TVs using quantum dots.

Your work is an example of the importance of basic research, of being curious, trying to solve mysteries without a particular endpoint or industrial application in sight. What message would you have for a young chemist starting out today working on such basic research?

The world is a huge place, and you could do basic research in a huge number of different areas. You want to pick a problem where, if you are spectacularly successful and you actually discover something really interesting, it might have some application in the world.

For better or for worse, you have to make a choice in the beginning, and it takes some intuition.

A good way to do it is you pick a subject that you know is important to technology, but there’s no understanding of the science at the present time. It’s a complete black box. Nobody understands the basic principles. That kind of problem, you can begin to take it apart and look to see what the basic steps are.

What changes for you now that you’ve won the Nobel Prize?

Well, this Nobel Prize, for better or for worse, has a special meaning in people’s minds all over the world. Yesterday when the mailman came I happened to be at the front door and he recognized me because my face was in the local newspaper. And he said, “I’ve never shaken the hand of a Nobel laureate before.”

For better or for worse, this is where I am right now, in a special category whether I like it or not. I still have my office in the university, but I don’t have a research group. I’m trying to leave that to the younger people. So this recognition probably means less for my research than it would if I was 40 years old.

I have received congratulations by email from a number of people who won the prize in past years. Their main recommendation is you must learn to say no. People will ask you to do all kinds of crazy things, and your time will be entirely taken up with these honorific university visits and giving little speeches. In order to have a real life and to be productive, you have to say no to all of these extraneous invitations.

And they also told me to have fun in Sweden! It’s an extremely elaborate schedule of events for that week in December when this award ceremony is. Extremely fancy. American culture, physics culture is different – if you win a prize from the American Physical Society, it’s a very low-key event. You just show up in an auditorium. It’s not even necessary to wear a suit.

So I will take my family, my grandchildren to Sweden and we’ll try to enjoy this as a great vacation.

Have you ever bitten into a nut or a piece of chocolate, expecting a smooth, rich taste, only to encounter an unexpected and unpleasant chalky or sour flavor? That taste is rancidity in action, and it affects pretty much every product in your pantry. Now artificial intelligence^[1] can help scientists tackle this issue more precisely and efficiently.

We’re a group of chemists who study ways to extend the life of food products, including those that go rancid. We recently published a study^[2] describing the advantages of AI tools to help keep oil and fat samples fresh for longer. Because oils and fats are common components in many food types, including chips, chocolate and nuts, the outcomes of the study could be broadly applied and even affect other areas, including cosmetics and pharmaceuticals.

Rancidity and antioxidants

Food goes rancid^[3] when it’s exposed to the air for a while – a process called oxidation. In fact, many common ingredients, but especially lipids^[4], which are fats and oils, react with oxygen. The presence of heat or UV light can accelerate the process.

Oxidation leads to the formation of smaller molecules such as ketones^[5], aldehydes^[6] and fatty acids^[7] that give rancid foods a characteristic rank, strong and metallic scent. Repeatedly consuming rancid foods can threaten your health^[8].

Fortunately, both nature and the food industry have an excellent shield against rancidity – antioxidants.

Antioxidants^[9] include a broad range of natural molecules, like vitamin C, and synthetic molecules capable of protecting your food from oxidation.

While there are a few ways antioxidants work^[10], overall they can neutralize many of the processes that cause rancidity and preserve the flavors and nutritional value of your food for longer. Most often, customers don’t even know they are consuming added antioxidants, as food manufacturers typically add them in small amounts during preparation.

But you can’t just sprinkle some vitamin C on your food and expect to see a preservative effect. Researchers have to carefully choose a specific set of antioxidants^[11] and precisely calculate the amount of each.

Combining antioxidants does not always strengthen their effect. In fact, there are cases in which using the wrong antioxidants, or mixing them with the wrong ratios, can decrease their protective effect – that’s called antagonism^[12]. Finding out which combinations work for which types of food requires many experiments, which are time-consuming, require specialized personnel and increase the food’s overall cost.

Exploring all possible combinations would require an enormous amount of time and resources, so researchers are stuck with a few mixtures that provide only some level of protection against rancidity. Here’s where AI comes into play.

A use for AI

You’ve probably seen AI tools like ChatGPT^[13] in the news or played around with them yourself. These types of systems can take in big sets of data^[14] and identify patterns, then generate an output that could be useful to the user.

As chemists, we wanted to teach an AI tool how to look for new combinations of antioxidants. For this, we selected a type of AI capable of working with textual representations^[15], which are written codes describing the chemical structure of each antioxidant. First, we fed our AI a list of about a million chemical reactions and taught the program some simple chemistry concepts, like how to identify important features of molecules.

Once the machine could recognize general chemical patterns, like how certain molecules react with each other, we fine-tuned it by teaching it some more advanced chemistry. For this step, our team used a database of almost 1,100 mixtures previously described in the research literature.

At this point, the AI could predict the effect of combining any set of two or three antioxidants in under a second. Its prediction aligned with the effect described in the literature 90% of the time.

But these predictions didn’t quite align with the experiments our team performed in the lab. In fact, we found that our AI was able to correctly predict only a few of the oxidation experiments we performed with real lard, which shows the complexities of transferring results from a computer to the lab.

Refining and enhancing

Luckily, AI models aren’t static tools with predefined yes and no pathways. They’re dynamic learners^[16], so our research team can continue feeding the model new data until it sharpens its predictive capabilities and can accurately predict the effect of each antioxidant combination. The more data the model gets, the more accurate it becomes, much like how humans grow through learning.

We found that adding about 200 examples from the lab enabled the AI to learn enough chemistry to predict the outcomes of the experiments performed by our team, with only a slight difference between the predicted and the real value.

A model like ours may be able to assist scientists developing better ways to preserve food by coming up with the best antioxidant combinations for the specific foods they’re working with, kind of like having a very clever assistant.

The project is now exploring more effective ways to train the AI model and looking for ways to further improve its predictive capabilities.