Dear Aventine Readers:

One day, you might hear the pilot on your flight announce that “AI has given us clearance to land.” But for now, there is great reluctance among experts and air traffic controllers to use AI in any safety-critical function. A nine-year experiment at London’s Heathrow Airport is a case study in how AI may — or may not — be integrated into aviation safety.

Also in this issue:

Thanks for reading, 

Danielle Mattoon
Executive Director, Aventine

Subscribe

Subscribe to our newsletter and be kept up to date on upcoming Aventine projects

When low clouds descend on London’s Heathrow Airport, it can cause chaos across Europe.

Heathrow is the continent’s busiest airport, with a combined 1,300 takeoffs and landings every day. When low clouds settle over the control tower, as they do often — this is London, after all — air traffic controllers can’t rely on straightforward visual checks to confirm whether departing planes have cleared the runway.

Instead, they depend heavily on radar and other instrument-based systems. That takes time and cuts capacity by around 20 percent, causing a ripple effect of delays, missed connections and canceled flights across Europe.

In the base of the control tower, there is a potential fix for this problem, one that could perhaps help other airports wrestling with poor visibility. For the past seven years, the UK’s National Air Traffic Control (known as NATS) has been testing an AI-powered image recognition tool to identify aircraft as they enter and leave the runway. NATS believes the new technology — which could act as eyes on the ground during times of low cloud cover — could get Heathrow operating close to full capacity by reducing delays caused by poor visibility.

The trick will be persuading air traffic controllers to actually use it. In theory, AI could relieve some of the enormous burden placed on controllers, but NATS is facing an uphill battle in convincing them, and some experts that AI can be used safely and effectively for aviation.

If the NATS technology is adopted, it would represent an unprecedented breakthrough in aviation. Despite AI tearing through industries, remaking companies and jobs at a feverish pace, there isn’t a single example of AI being used in a safety-critical aviation task anywhere in the world. The reluctance is largely due to the distrust air traffic controllers and authorities have for a black-box technology like AI; this is a contrast with carmakers, who are pushing to get autonomous vehicles on the road. For those in the safety-obsessed aviation business, there’s an almost impossibly high bar for trusting current AI systems to anticipate the kind of unpredictable situations — edge cases — that air traffic controllers must face routinely, no matter how large the dataset and how exhaustively the model is trained.

At the same time, air traffic control seems like a field that is ripe for AI augmentation. Currently, the profession is woefully understaffed. A study found that in 2021, staffing shortages were responsible for about one-quarter of all flight delays in Europe. In the US, 90 percent of all air traffic control centers operate below the staffing levels recommended by the Federal Aviation Administration. Candidates must pass rigorous training for high-level cognitive tasks like processing and sharing new information, multitasking, vigilance, working memory and visual-spatial processing. And recruiting new controllers is difficult.

But the new technology could address more than just inconvenience and lost revenue. There have been minor runway collisions at airports around the world in the past few years, including one at Heathrow in 2024. While no one has died in recent collisions at Heathrow, the single deadliest crash in aviation history was the runway collision of two passenger jets in dense fog in Tenerife, Spain, in 1977 that killed 583 people.

Heathrow’s experiment with AI began in 2017 when NATS made a major investment in Searidge Technologies, an aviation software company, to set up what they called a “digital tower” — a system that coordinates data from an airport, including images from cameras. The company began by installing 20 ultra-high-definition cameras around the airport. The images from these cameras were then fed into “Aimee,” a computer visual processing model that learned to identify the different planes and other vehicles at the airport. In a three-month trial, Aimee analyzed 40,000 arrivals, achieving an accuracy rate for identifying the correct aircraft of over 98 percent.

“We need a system that can recognise different aircraft types in all conditions, but especially at times of low visibility,” said Marco Rueckert, Searidge’s chief technology officer. “Using Aimee to process and interpret visual data is the most cost-effective way of doing that with confidence.”

The goal is to integrate Aimee’s input on an air traffic controller’s screen with either a color or symbol change to indicate when a plane has left the runway. Rueckert stressed that Aimee is not designed to replace the controller’s judgment, but to supplement it. Controllers would still verify Aimee’s input with a visual check or radar.

Seabridge finished its initial trial in 2019 and began talks with Heathrow about investing in new cameras for further tests. Then Covid hit, devastating air travel, and the talks were put on hold. As a result, Aimee hasn’t been installed in Heathrow’s air traffic system or tested by the controllers there. In the meantime, the company has been conducting tests at Singapore’s Changi Airport and is trying to convince another UK airport to test AI technology on the runways. The company has also been using AI to find efficiencies in noncritical functions at Heathrow by monitoring how planes and other vehicles move across the tarmac, checking on the number of planes in a departing flight queue and gauging how much time planes spend at the gates. Ultimately the goal would be to channel all of this information into a system that would help the entire airport operate more smoothly.

But Searidge faces tough odds. For starters, the company will need to convince Heathrow to spend several million pounds upgrading its runway cameras in order to further test the technology.

The bigger hurdle would be regulatory. Searidge would then need to get approval from the UK regulator, the Civil Aviation Authority (CAA). Given that no AI systems are currently in use in aviation safety, approval is not a given.

Helena Sjöström Falk spent 37 years working as an air traffic controller in Sweden before becoming president of the International Federation of Air Traffic Controllers. She said that air traffic controllers are not averse to AI but want it to be adopted cautiously.

“We have to recognize that there will be developments, but the skepticism is certainly there,” she said. “It’s all too easy to say ‘this is going to be amazing and fantastic’ when you’re not the one that has to deal with it. We have a bit more of a pragmatic outlook. Let’s take it slow. Let’s make sure that things are safe first.”

There is an inherent risk that AI systems can turn human operators into passive observers. This risk is especially concerning in air travel because air traffic controllers need to maintain constant vigilance.

“Every day, in front of the radar or in the tower, we are constantly looking at things and checking and checking again and thinking, ‘what if? What if? What if?’” Sjöström Falk said. “I have to have a Plan B, Plan C, Plan D. That is our job. We are skeptical because we are the end users. We are the ones dealing with the safety of the air traffic, the passengers and the pilots.”

“Is AI going to take over air traffic control?” she asked. “Not in my lifetime. Eventually yes, but it’s going to take a very, very, very long time,” she said. “Is it going to be a great addition to our toolbox? Yes, definitely. It’s a slow process, like everything in aviation.”

Human eggs can be rejuvenated, which could improve IVF. An important age-related defect that causes genetic errors in embryos may be reversible, according to new research led by scientists at the Max Planck Institute for Multidisciplinary Sciences in Göttingen. The work, first reported by The Guardian, suggests that adding a protein back into aging human eggs may restore a crucial step in reproduction. The intervention targets meiosis, the process by which an egg cell discards half of its genetic material to make room for the half contributed by sperm. In older eggs, that process often goes wrong, increasing the risk that embryos end up with the wrong number of chromosomes, resulting in miscarriage or developmental disorders. The researchers found that levels of a protein called Shugoshin 1 decline as eggs age. In experiments with both mouse and human eggs, they used microinjections to restore the protein and saw the meiosis defect partially reversed. In human eggs, the proportion showing abnormal meiosis fell from 53 percent in untreated cells to 29 percent in treated ones. The results were posted as a preprint on bioRxiv and have not yet been peer reviewed. A startup founded by the team, Ovo Labs, seeks to commercialize the approach and is reportedly in discussions with regulators about launching a clinical trial. If the findings hold up, the technique could one day improve IVF success rates for older patients. 

Engineered diamonds are ultra-sensitive sensors. The natural diamond industry is in turmoil as lab-grown gems drive prices down. But that same manufacturing technology is turning out to be a major scientific asset. As the Financial Times reports, researchers are now creating nano-engineered diamonds that function as extraordinarily sensitive sensors. These so-called “technology diamonds” are produced by deliberately introducing tiny defects into the diamond lattice. In place of two carbon atoms, scientists insert a nitrogen atom and leave an adjacent void, creating what’s known as a nitrogen-vacancy center. When exposed to magnetic fields, these defects emit light; changes in the field alter its brightness. The result is a sensor so responsive it can reportedly detect a car driving down a road 100 meters away. That sensitivity opens the door to a wide range of applications: ultra-precise navigation systems that read the Earth’s magnetic field; geological prospecting tools that spot subtle variations linked to mineral deposits; and even medical devices capable of measuring the tiny magnetic signals produced by heartbeats, potentially replacing electrocardiograms. While diamond-based sensors have been under development for more than a decade, progress has accelerated recently. In 2022, a diamond sensor system took up half a room; today, the hardware is the size of a smartphone, and the crystals can be produced reliably enough to cost just a few thousand dollars. That drop in price may help them power a new generation of scientific instruments.

A new pill could tell doctors if you took your meds. Medication nonadherence is a huge problem. It contributes to an estimated 125,000 preventable deaths each year in the US, and nearly half of patients with chronic conditions stop taking new prescriptions within the first year. Researchers at MIT now report a potential solution: a pill that can signal when it has been swallowed, before almost completely dissolving in the body. Described in Nature Communications, the device uses a tiny biodegradable antenna attached to a radio-frequency chip less than half a millimeter across. The components are encased in a coating embedded with metallic particles, which blocks any signal until the pill reaches the stomach and the coating breaks down. Once that happens, an external receiver can detect the signal, confirming that the pill has been ingested. All of the components except the microscopic chip dissolve, leaving a particle small enough to easily pass through the digestive system. The design is compatible with standard capsules, meaning it could, in principle, be added to many existing medications. In tests in pigs, whose digestive systems are similar to those of humans, the signal was successfully detected from inside the stomach at distances of up to two feet away. The researchers stress that the technology is not intended for routine use, but rather for cases where missing a dose, even by hours, could be dangerous or catastrophic.

AI materials discovery now needs to move into the real world, from MIT Technology Review
4,000 words, or about 16 minutes

In late 2023, Google’s AI lab DeepMind made a bold claim: It had used artificial intelligence to discover “millions of new materials.” The materials science community was underwhelmed. There were few genuinely novel compounds in the dataset, and all of them were theoretical — not demonstrated to be manufacturable, stable, or useful. But this MIT Technology Review story describes how the field is now maturing. Instead of generating lists of hypothetical materials, a new wave of startups, including Lila Sciences and Periodic Labs, is trying to close the loop between prediction and production. Their goal is not just to propose new materials, but to work out how to make them, test their properties and refine the recipes using AI as a guide at every step. There is still a long road ahead. Discovering genuinely useful materials is notoriously difficult, and commercializing them is harder still. But materials science has been relatively stagnant for decades, and these well-funded, AI-wielding companies may represent the best chance in recent history to accelerate progress in everything from batteries and catalysts to semiconductors and clean energy.

The man behind the fall of offshore wind, from Canary Media
4,500, or about 18 minutes

No, it’s not Donald Trump. This Canary Media profile focuses on David Stevenson, a man who has solar panels on his roof, drives a hybrid car and once co-founded the Delaware Green Building Council to promote energy efficiency. He is not a climate skeptic. But over the past decade, he has become one of the most influential opponents of offshore wind power in the United States. After a career in construction, Stevenson entered the world of think tanks in 2010. By 2017, he had developed a strong opposition to offshore wind, largely on cost grounds. A few years later, as some members of the political right began claiming that offshore wind projects were responsible for whale deaths along the US coast, Stevenson did not endorse the theory — but, according to this reporting, was happy to let it spread. A 2025 study by researchers at Brown University found Stevenson at the center of a network of activists and political operators lobbying against offshore wind, backed by funding from the fossil fuel industry. The result has been delays, legal challenges and growing public opposition to projects that were once seen as politically uncontroversial. And this story gets at a larger point: Outright denial of climate change appears to be softening, but in its place is a growing trend of powerful individuals blocking specific climate-related projects they happen to dislike.

90 Minutes to Give Baby Luna a New Heart, from The New York Times
3,500 words, or about 14 minutes

Some medical interventions carry unimaginable stakes. The first time a surgeon performs an infant heart transplant is one of them. This New York Times story follows Dr. Maureen McKiernan as she performs her first such operation at New York-Presbyterian Morgan Stanley Children’s Hospital in northern Manhattan. Infant heart transplants are rare — performed around 100 times a year across the entire United States — and they lie at the limits of what modern medicine can achieve. In a race against time, a heart must be removed from a child who has died, while another baby’s failing heart is taken out; surgeons then painstakingly reconnect tiny veins and arteries and attempt to bring the new heart to life. The details are at once fascinating and terrifying: a donor heart the size of a strawberry; sutures finer than a human hair; blood vessels so thin they resemble paper. Yet against those odds, the operations succeed. And perhaps the best detail of all is that the procedure succeeded for Baby Luna, the recipient of the donor heart in this story, who was eventually strong enough to leave the hospital and make it home in time for Christmas.