Nearly 4 in 10 American adults reported engaging in dangerous cleaning practices to prevent COVID-19, such as washing food with bleach, using household disinfecting products on their skin or intentionally inhaling vapors from cleaning products, according to a recent survey.
A report from the Centers for Disease Control and Prevention (CDC) follows an earlier study describing an increase in calls to poison control centers regarding exposure to household cleaners, during the COVID-19 pandemic.
Technion Professor Ido Kaminer and his team have made a dramatic breakthrough in the field of quantum science: a quantum microscope that records the flow of light, enabling the direct observation of light trapped inside a photonic crystal.
Their research, “Coherent Interaction Between Free Electrons and a Photonic Cavity,” was published in Nature. All the experiments were performed using a unique ultrafast transmission electron microscope at the Technion-Israel Institute of Technology. The microscope is the latest and most versatile of a handful that exist in the scientific world.
“We have developed an electron microscope that produces, what is in many respects, the best near- field optical microscopy in the world. Using our microscope, we can change the color and angle of light that illuminates any sample of nano materials and map their interactions with electrons, as we demonstrated with photonic crystals,” explained Prof. Kaminer. “This is the first time we can actually see the dynamics of light while it is trapped in nano materials, rather than relying on computer simulations,” added Dr. Kangpeng Wang, a postdoc in the group and first author on the paper.
A team including researchers from the Department of Chemistry at the University of Tokyo has successfully captured video of single molecules in motion at 1,600 frames per second. This is 100 times faster than previous experiments of this nature. They accomplished this by combining a powerful electron microscope with a highly sensitive camera and advanced image processing. This method could aid many areas of nanoscale research.
When it comes to film and video, the number of images captured or displayed every second is known as the frames per second or fps. If video is captured at high fps but displayed at lower fps, the effect is a smooth slowing down of motion which allows you to perceive otherwise inaccessible details. For reference, films shown at cinemas have usually been displayed at 24 frames per second for well over 100 years. In the last decade or so, special microscopes and cameras have allowed researchers to capture atomic-scale events at about 16 fps. But a new technique has increased this to a staggering 1,600 fps.
Paleontologists at St Petersburg University created the most detailed virtual 3D-model of the endocranial cast and blood vessels of the head of an ankylosaurian.
Paleontologists from St Petersburg University have been the first to study in detail the structure of the brain and blood vessels in the skull of the ankylosaur Bissektipelta archibaldi. It was a herbivorous dinosaur somewhat similar in appearance to a modern armadillo. The first three-dimensional computer reconstruction of a dinosaur endocast made in Russia — a digital cast of its braincase — was of help to the scientists. It made it possible to find out that ankylosaurs, and Bissektipelta in particular, were capable of cooling their brains, had an extremely developed sense of smell, and heard low-frequency sounds. However, their brain was one and a half times smaller than that of modern animals of the same size.
Ankylosaurs appeared on Earth in the middle of the Jurassic — about 160 million years ago — and existed until the end of the dinosaur era, which ended 65 million years ago. These herbivorous animals were somewhat reminiscent of modern turtles or armadillos, were covered with thick armor, and sometimes even had a bony club on the tail. The researchers became interested in the uniquely-preserved remains of ankylosaurs from Uzbekistan. Although these fossils have been known for 20 years, only now have the scientists had a unique opportunity to study the specimens from the inside using cutting-edge methods.
Over the last few years, the size of deep learning models has increased at an exponential pace (famously among language models):
And in fact, this chart is out of date. As of this month, OpenAI has announced GPT-3, which is a 175 billion parameter model—or roughly ten times the height of this chart.
As models grow larger, they introduce new infrastructure challenges. For my colleagues and I building Cortex (open source model serving infrastructure), these challenges are front and center, especially as the number of users deploying large models to production increases.