The clues were hiding in plain sight for four hundred years - microscopic ridges of dried oil paint, each one a fingerprint left at a crime scene nobody knew they were investigating. The suspect: El Greco's own workshop, long accused of finishing The Baptism of Christ after the master died in 1614....
Back in 2012, AlexNet made computers weirdly good at recognizing images, but it left one giant hole in the plot: spotting trouble in a picture is not the same as getting an actual patient to walk through the right clinic door afterward. That missing piece is the mystery at the center of this new...
3 reasons this paper matters, starting with the least obvious.
If your battery were a house, the graphite anode would be the foundation - and for the past thirty years, scientists have been living in it without fully understanding why the basement floods every time they rearrange the furniture. The old approach to modeling how lithium ions nestle between...
That is the basic plot of a new npj Digital Medicine study on Xuanwu-NeuroAid, a domain-specific large language model built for emergency neurological diagnosis. In a prospective shadow evaluation of 433 patients, the model hit 79.4% diagnostic accuracy, while emergency physicians landed at 65.4%....
"These brain-scan AI papers are just glorified age detectors," says the standard criticism, usually while everyone nods like they have personally audited 2,000 MRI volumes. Fair complaint. Mild cognitive impairment, or MCI, often hides inside the same structural brain changes that show up in...
Guess how many new ingredients a machine learning liver test needs to beat an old standby like FIB-4. Ten? Twenty? A whole smoothie of biomarkers? Wrong. This paper says two added metabolites - tyrosine and taurocholic acid - might be enough to make the usual fibrosis workup look a little...
Biomedical imaging has an annoying habit of asking for everything at once: go deeper, go faster, use less light, and please do not fry the sample. According to Yuanjie Gu and colleagues, that bargain usually ends with noisy 3D volumes and missing structure, especially in optical coherence...
Most people assume the model with the bigger score wins. More AUC, more confetti, ship it to the clinic, everybody go home. This new paper says that instinct is exactly how you end up with a very confident spreadsheet and a very confused doctor.
A team of researchers recently fed 9,555 natural compounds into an AI screening pipeline and out popped a molecule from the ginkgo tree that appears to fix broken cellular garbage trucks in motor neurons. Standard Tuesday, really.
At 2:13 a.m. in a sleep lab, a tech is staring at a wall of squiggly EEG lines, oxygen drops, chest bands, and enough overnight data to make a spreadsheet tap out. This is where the paper by Sharafkhaneh and colleagues walks onto the job site and says: maybe the humans should not have to hand-sort...
Where Li et al. (2018) eyeballed China's lake CO₂ output at a hefty 15.98 Tg C per year, Gao et al. (2023) trimmed that estimate down to a leaner 8.07 Tg C per year with seasonal corrections, and now Feng et al. (2026) have entered the ring with the most dynamic model yet - one that tracks how...
Google DeepMind unleashed GNoME and predicted 2.2 million new crystal structures through sheer brute-force deep learning, essentially throwing a massive neural network at the periodic table and seeing what stuck (Merchant et al., Nature, 2023). Meanwhile, a team led by Yongxin Lyu and collaborators...
Guess what percentage of human proteins have drug-friendly binding pockets that show up in a standard crystal structure. If you said "most of them," congratulations - you're wrong, and that wrong answer has haunted pharmaceutical companies for decades. Roughly 85% of the human proteome has been...
That statement is rude, slightly unfair, and only half wrong. The paper here is not claiming humans arrived with a magical bonus feature pack. It is saying some of our DNA regulatory regions seem easier to open, read, and use than the matching regions in other primates. Nit: "special" is a bad...
OK so let me give you the 30-second version before we get into the weeds. Density functional theory - DFT for short - is the workhorse of computational chemistry. It's the method behind basically every simulation that predicts how molecules behave, from designing new drugs to engineering better...
Most chemistry papers feel like careful roadwork. This one shows up with a transit map and says, actually, what if the soup could run a recurrent neural network. Very normal week in science.
The classical AI model could feel itself losing grip. Three timesteps into a turbulent flow prediction, its confidence was already taking on water - outputs drifting, small-scale features dissolving into numerical mush, the forecast equivalent of a ship's compass spinning wildly in a magnetic...
Taking a deterministic weather model, subtracting its predictions from reality to isolate the "residual chaos," and then training a generative model on that chaos alone - it sounds like the kind of plan you sketch on a napkin at 2am and expect to quietly abandon by morning. Except Couairon et al....
A fully printed, bendable artificial brain synapse just hit 93.91% image recognition accuracy - and it's made from ink.