Researchers engineered a metal structure that can’t be sunk, a data storage technology that can’t be destroyed and an artificial digital retina that doesn’t damage the eye. This week’s coolest scientific discoveries present a wide world of possibilities.
What is it? Learning from “diving bell spiders and rafts of fire ants,” researchers at the University of Rochester in New York have created a metallic structure so hydrophobic that it won’t sink in water — it floats even when holes are drilled in it.
Why does it matter? The most dramatic possibility is unsinkable ships, but Chunlei Guo, a Rochester professor of optics and physics, said other uses could include a flotation device that’s buoyant even if damaged or electronic monitoring devices that endure the ravages of long exposure to the sea. Guo is the co-author of a paper describing the design in ACS Applied Materials and Interfaces.
How does it work? The new findings build on an earlier development out of Guo’s lab. Emulating nature, the researchers used lasers to create micro- and nanoscale patterns on the surfaces of metal, trapping air and rendering them superhydrophobic — that is, extremely water-repellant. In the latest advance, the researchers placed two aluminum discs parallel at a precise distance from one another, with their hydrophobic surfaces facing inward. The surfaces repelling the water created a kind of air bubble between the discs, rendering it buoyant. Because the important parts face inward, moreover, they remain free from wear and tear.
What is it? How good can artificial intelligence be at thinking through complex design problems? “The answer is quite good,” said Jonathan Cagan, part of the team at Carnegie Mellon University’s College of Engineering that trained AI to help engineer a bridge.
Why does it matter? “Big design problems require creative and exploratory decision-making, a skill in which humans excel,” according to a release from CMU. AI, by contrast, excels at thinking through problems within a defined set of rules. The CMU researchers wondered whether they could help it be a little more creative. They taught the AI to learn by watching humans, which is different from “just mimicking or regurgitating solutions that already exist,” Cagan said. “It’s learning how people solve a specific type of problem and creating new design solutions from scratch.”
How does it work? The engineers chose the complex problem of how the trusses of bridges work to form a complete structure. Deep neural networks working together in what’s called a prediction-based situation observed “a set of five sequential images” in the progression of a bridge design. The AI agents were then asked to predict what the design would be based on the previous configurations. The AI agents did well, but the researchers who programmed them imagine that their work will augment rather than supplant that of human engineers. While the AI could come up with effective designs, they were working only on the basis of visual information and not other aspects that engineers must consider such as the weight of the materials. The research is described further in the Journal of Mechanical Design.
What is it? A collaboration between Microsoft and Warner Bros. has yielded a glass disc the size of a drink coaster that contains the classic Christopher Reeve movie “Superman,” and it will work even after being “boiled in hot water, baked in an oven, microwaved, flooded, scoured, demagnetized,” according to Microsoft.
Why does it matter? The technique used to produce this disc could be used to protect other cultural artifacts against destruction — a concern for Warner Bros., which has been seeking ways to preserve its vast archive that are stable over time and don’t require specific temperatures and constant monitoring. But it could have broader utility: humans produce more data, and there’s a need for better, and cheaper, storage technologies. As Microsoft points out, hard drives and magnetic tape wear out, and file formats become obsolete. Tech and media companies have been searching for a storage format that lasts.
How does it work? Enter glass storage — part of a Microsoft Research endeavor called Project Silica, which combines laser optics and artificial intelligence. According to Microsoft, “A laser encodes data in glass by creating layers of three-dimensional nanoscale gratings and deformations at various depths and angles. Machine learning algorithms read the data back by decoding images and patterns that are created as polarized light shines through the glass.” The proof was the “Superman” disc, which is 75 by 75 millimeters and 2 millimeters thick. It’s not faster than a speeding bullet, but it could last for centuries.
What is it? Researchers at Stanford University have taken a step toward building an artificial digital retina that could help restore sight to blind people by surmounting a big obstacle: heat.
Why does it matter? Think of when you go to a website with a lot of pop-up ads and autoplay videos on it. If you linger on the page, your computer heats up while trying to process all that data, indicated by the whirring of the internal fan. Researchers trying to create an artificial digital retina have encountered a similar problem. The typical device is a tiny computer chip with electrodes poking out that records neuron activity, then transmits visual information from a camera to the brain. The eye collects so much visual data, though, that the electronics get too hot, which could burn eye tissue.
How does it work? As Stanford’s Andrew Myers explained, “To convey visual information, neurons in the retina send electrical impulses, known as spikes, to the brain. The problem is that the digital retina needs to record and decode those spikes to understand the properties of the neurons, but that generates a lot of heat in the digitization process.” The Stanford researchers found a way to separate the signals from the visual noise by weeding out less essential visual information and focusing on truly important data. Their technology (described further in IEEE Transactions on Biomedical Circuits and Systems) then digitizes only the unique spikes to send onward to the brain, keeping the electronics cool in the process.
What is it? Researchers from the City University of Hong Kong and the Karolinska Institutet in Stockholm developed a protein that they say can improve the accuracy of CRISPR-Cas9 gene editing.
Why does it matter? CRISPR-Cas9 is a promising “cut-and-paste” bioengineering technology with the potential to treat medical problems from cancer to blood disorders by simplifying gene editing. Rather than introduce new copies of a gene into cells, it’s essentially able to make repairs on-site. But it has some problems with accuracy, with potentially dire consequences, as a release from City University of Hong Kong notes: “Unintended modifications of the genomes could potentially lead to serious consequences, such as cancers.” The new development improves its precision.
How does it work? Researchers rely today on two Cas9 variants, called SpCas9 and SaCas9. They’ve been able to engineer SpCas9 to increase its precision, but the enzyme is still too large for the delivery method commonly used in gene therapy. The Hong Kong team focused on the much smaller enzyme SaCas9, finding a way to increase its accuracy. The engineered enzyme “reduced the off-target activity by about 90%” compared with its unmodified counterpart — what’s called the wild-type SaCas9. The study was published in the Proceedings of the National Academy of Sciences of the United States of America.