Large and unusual rock formations could help seismologists forecast earthquakes, an ambitious European climate-modeling project will create a “digital twin” of the entire Earth, and a tiny nanogenerator could be scaled up into a sizable machine that “scavenges” wind energy. Scientists are thinking big in this week’s coolest developments.
What is it? Princeton University researchers developed a tool to help address issues of bias in artificial intelligence — specifically, the ways that AI learns from sets of images.
Why does it matter? Images are used to train AI systems to do an increasing number of tasks, according to Princeton, “from credit services to courtroom sentencing programs.” Such images are typically scraped from the internet; the AI learns to detect patterns after processing huge amounts of visual information. But because it’s processing information from the human world, AI is also susceptible to replicating human biases, including racial, gender and geographical prejudices.
How does it work? One such bias flagged by the Princeton system — an open-source tool called REvealing VIsual biaSEs, or REVISE — relates to images of people, flowers and gender: It noticed that men appeared in photos with flowers more “in ceremonies or meetings,” whereas women “tended to appear in staged settings or paintings.” It also was able to flag a discrepancy among images annotated as “dish”: In images from East Asia that word tended to refer to food, whereas in other parts of the world it more often referred to satellite dishes or plates. The study describing the tool was presented in August at the (virtual) European Conference on Computer Vision.
What is it? You know when you see a big boulder resting precariously atop another, smaller boulder? Turns out geologists have a term for that: precariously balanced rocks, or PBRs. And a new study from Imperial College London explains how PBRs might help forecast earthquakes.
Why does it matter? “Our new approach could help us work out which areas are most likely to experience a major earthquake,” said engineering professor Anna Rood, lead author of a new study in AGU Advances. Beyond its obvious import to communities that may be situated in riskier areas, that kind of information can be useful for civil engineers as they design bridges, dams and other infrastructure, and could help set insurance prices in earthquake-prone locales.
How does it work? Seismologists commonly look to nearby fault lines and recent history when trying to determine earthquake hazard in a given area. But they are limited in their understanding of large, rare quakes that may have occurred in the distant past. Focusing their study on PBRs near a nuclear plant in California, Rood and her team used a dating technique to determine how long the rocks had been in their current, precarious location, then used digital modeling to determine how much earth shaking would have to occur in order to knock the rocks over. Basically, that lets them know “the upper limit of earthquake shaking that has occurred since they were first formed,” according to a release from Imperial College London — and could thereby “boost the precision of hazard estimates for large earthquakes by up to 49%.”
What is it? Digital twin technology is all the rage — companies including GE use it, for instance, to create computer-simulated models of machines, allowing them to help predict wear and tear and schedule maintenance. Now the European Union is taking that general principle and applying it to a much, much, much larger object: the Earth itself.
Why does it matter? A digital twin of the entire planet would allow researchers to forecast climatic events like droughts, floods and fires. “It’s a really bold mission,” U.S. Department of Energy climate scientist Ruby Leung told Science. That digital simulation, which the EU is calling Destination Earth, will also allow policymakers to see the effects of climate change, and to model the impacts of measures taken to slow it.
How does it work? The overall cloud-based “platform” of Destination Earth will include “digital replicas of various aspects of the Earth system,” per the EU, including ocean circulation, food security and weather forecasting — and many other planetary subsystems. One goal will be to capture the planet’s atmosphere at a resolution of 1 kilometer, which will enable researchers to unlock the “third dimension” of climate modeling, as another researcher told Science: convection. The project will be implemented over the next decade, starting in 2021.
What is it? The pharmaceutical company Regeneron released promising results from trials of an antibody cocktail that treats COVID-19.
Why does it matter? The cocktail is called REGN-COV2, and the trial — the initial phase of which included 275 people — is part of a larger program to test how the cocktail works both as a treatment for people hospitalized with COVID-19 infection and as a way to prevent infection in people who have been exposed.
How does it work? Early results suggest that the cocktail worked to rapidly reduce viral load and associated symptoms in infected patients, and also suggest that the treatment is most effective in “patients who had not mounted their own effective immune response,” explained Regeneron president George D. Yancopoulos. “These patients were less likely to clear the virus on their own, and were at greater risk for prolonged symptoms.” Jeanne Marrazzo, an infectious disease specialist at the University of Alabama at Birmingham, told CNN that a treatment that reduces the amount of virus in people’s throats could also theoretically make them less infectious to others.
What is it? Some wind turbine designers go as big as possible, and some go in another direction altogether: Researchers in China have created a tiny device that can “scavenge wind energy from breezes as little as those created by a brisk walk.”
Why does it matter? In truth, the device is not technically a turbine — it is a “nanogenerator” that exploits a phenomenon called the triboelectric effect. Ya Yang — a researcher at the Chinese Academy of Sciences’ Beijing Institute of Nanoenergy and Nanosystems, and the senior author of a new paper in Cell Reports Physical Science — says his hopes are twofold: first, to power small electronic devices like phones. More ambitiously, though, he wants to massively scale up the technology to provide a low-cost alternative to wind turbines that could be put in places, like cities, where conventional wind turbines may not be feasible. “I’m hoping to scale up the device to produce 1,000 watts, so it’s competitive with traditional wind turbines,” Yang said.
How does it work? The centerpiece of the nanogenerator couldn’t be simpler: It’s two strips of plastic that flutter when air blows through them. They become electrically charged the same way that rubbing a balloon on your head creates static electricity; the triboelectric effect refers to a type of electrification that happens when materials become charged after coming into contact then being separated. The device that Yang and his colleagues created is capable of powering 100 LED lights and temperature sensors.