The new science of “emergent misalignment” explores how PG-13 training data — insecure code, superstitious numbers or even extreme-sports advice — can open the door to AI’s dark side.
Get the latest international news and world events from around the world.
Skyrmions as Active Matter
Pairs of skyrmions—tiny whirlpools that emerge in some magnetic materials—might be able to self-propel, a behavior reminiscent of that of active-matter systems such as motile bacteria.
In nature, the collective motion of birds and fish can generate impressive dynamics and unique structures, as seen in flocks of starlings and shoals of sardines. The science of active matter studies such complex behaviors across a wide range of scales and origins [1], and it has attracted growing interest over the past three decades. Active matter encompasses not only living things but also inanimate objects. Examples include active colloids [2] and active liquid crystals [3] that are able to self-propel—that is, move by themselves powered by internal energy sources. Now Clécio de Souza Silva and colleagues at the Federal University of Pernambuco in Brazil have suggested an intriguing addition to the active-matter catalog: coupled pairs of skyrmions, whirlpool-like spin arrangements that emerge in certain magnetic materials.
Rapid-response protocol promises to reveal supernovae only hours after they explode
“They shine thanks to nuclear fusion in their cores, but once the star has burned through progressively heavier atoms—right up to the point where further fusion no longer yields energy—the core collapses. At that point, the star collapses because gravity is no longer counterbalanced; the rapid contraction raises the internal pressure dramatically and triggers the explosion.”
The first hours and days after the blast preserve direct clues to the progenitor system—information that helps distinguish competing explosion models, estimate critical parameters, and study the local environment. “The sooner we see them, the better,” Galbany notes.
Historically, obtaining such early data was difficult because most supernovae were discovered days or weeks after the explosion. Modern wide-field, high-cadence surveys—covering large swaths of sky and revisiting them frequently—are changing that picture and allowing discoveries within mere hours or days.
Excavating Eridu: Observations explore nature of massive ancient galaxy
By analyzing the data from the James Webb Space Telescope (JWST) and the Hubble Space Telescope (HST), astronomers from the University of Wisconsin-Madison and elsewhere have probed the properties of a massive and old galaxy designated SMILES-GS-191748. Results of the study, published August 7 on the pre-print server arXiv, shed more light on the nature of this galaxy.
SMILES-GS-191748 is a massive and quiescent galaxy at a redshift of 2.675. The galaxy most likely contains a very old stellar population that first formed when the universe was young.
Given that very little is known about the properties of SMILES-GS-191748, a team of astronomers led by University of Wisconsin-Madison’s Ian McConachie decided to inspect this galaxy using JWST and HST. They nicknamed SMILES-GS-191748 “Eridu,” after the ancient Bronze Age Sumerian city in Mesopotamia due to the galaxy’s suspected early formation time and apparent quiescent nature.
Discovery of hidden faults sheds light on mystery of ‘slow earthquakes’
Scientists have uncovered a key piece of the puzzle behind the unusual “slow earthquakes” occurring off the east coast of New Zealand’s North Island.
A new international study, published in Science Advances, identifies hidden fault structures called polygonal fault systems (PFSs) as a major influence on the behavior of the northern Hikurangi subduction zone.
These shallow geological features, found in sediments entering the subduction zone, appear to play a critical role in where and how slow slip earthquakes occur.
Self-powered photodetector achieves 20-fold sensitivity boost using novel device structure
Silicon semiconductors used in existing photodetectors have low light responsivity, and the two-dimensional semiconductor MoS₂ (molybdenum disulfide) is so thin that doping processes to control its electrical properties are difficult, limiting the realization of high-performance photodetectors.
A KAIST research team has overcome this technical limitation and developed the world’s highest-performing self-powered photodetector, which operates without electricity in environments with a light source. This paves the way for precise sensing without batteries in wearable devices, biosignal monitoring, IoT devices, autonomous vehicles, and robots, as long as a light source is present.
Professor Kayoung Lee’s research team from the School of Electrical Engineering developed the self-powered photodetector, which demonstrated a sensitivity up to 20 times higher than existing products, marking the highest performance level among comparable technologies reported to date. The work is published in the journal Advanced Functional Materials.
Unique fingerprints in 3D printing may foil adversaries
3D printing is a simple way to create custom tools, replacement pieces and other helpful objects, but it is also being used to create untraceable firearms, such as ghost guns, like the one implicated in the late 2024 killing of UnitedHealthcare CEO Brian Thompson.
Netanel Raviv, assistant professor of computer science & engineering in the McKelvey School of Engineering at Washington University in St. Louis, led a team from the departments of Computer Science & Engineering and Biomedical Engineering that has developed a way to create an embedded fingerprint in 3D-printed parts that would withstand the item being broken, allowing authorities to gain information for forensic investigation, such as the identity of the printer or the person who owns it and the time and place of printing.
The research will be presented at the USENIX Security Symposium Aug. 13–15, 2025, in Seattle. The first authors of the paper are Canran Wang and Jinweng Wang, who earned doctorates in computer science in 2024 and 2025, respectively. The research is published on the arXiv preprint server.
Carbon nanotube ‘smart windows’ offer energy savings by modulating near-infrared light transmission
Half of the sun’s radiant energy falls outside of the visible spectrum. On a cold day, this extra infrared light provides additional warmth to residential and commercial buildings. On a warm day, it leads to unwanted heating that must be dealt with through energy-intensive climate control methods such as air-conditioning.