Researchers at Technion-Israel Institute of Technology and the Peng Cheng Laboratory have recently created a new neuromorphic computing system supporting deep belief neural networks (DBNs), a generative and graphical class of deep learning models. This model is based on silicon based memristors, which are energy-efficient devices that can both store and process information. The artificial synapses were fabricated using commercial complementary-metal-oxide-semiconductor (CMOS) processes. These silicon based synapses have numerous advantageous features, including analogue tunability, high endurance, long retention time, predictable cycling degradation, and moderate variability across different devices. The results showed over 97% accurate recognition of handwritten digits when using Y-Flash based memristors. Memristors measured in a probe station (Source: https://techxplore.com)
AI system learns concepts shared across video, audio, and text
Visual learning is considered a highly efficient way to understand a concept. Researchers at the Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed an artificial intelligence (AI) technique that allows machines to learn concepts shared between different modalities, such as videos, audio clips, and images. The model learns to represent data in a way that captures concepts which are shared between visual and auditory modalities. The work focuses on representation learning, which is a form of machine learning that seeks to transform input data to make it easier to perform a task like classification or prediction. The model still has some limitations that need to be addressed. For one, their research focused on data from two modalities at a time, but in the real world humans encounter many data modalities simultaneously.
Design for AI engineering framework developed
Researchers from Carnegie Mellon University have devel- to educate and encourage the academic and industrial engineering community to adopt AI engineering design. The team outlines three personas as necessities for DfAI: engineering designers, design repository curators, and AI developers. An engineering designer may be a person or team responsible for developing the specifications of a new project. They are the problem solvers who can understand the engineering constraints as well as the AI algorithms. Design repository curators must take the role of a database maintainer one step further by having engineering design and manufacturing knowledge to deliver design engineers the data management tools to meet workflow demands and be extensible to future demands. Finally, the AI developers must be able to ideate, develop, market, and continuously improve AI software products to help the design engineers.
A watch that works only when attached organism is healthy
life—literally. They used the electrically conductive single-cell organism known as “slime mold” to create a watch that works only when the organism is healthy, requiring the user to provide it with food and care. The organism is placed in an enclosure on the watch, and the user must regularly feed it a mixture of water and oats to induce its growth. When it reaches the other side of the enclosure, it forms an electrical circuit that activates the heart rate monitor function. When tested, researchers found a high level of attachment to the watch, with some users saying it felt like a pet—even naming it or putting their partner in charge of the feeding when they got sick. This also depicted the users’ emotional response when they were asked to neglect the organism, expressing guilt or even grief.
Fusion reaction generates more energy than used to create it
A team of researchers at the National Ignition Facility (NIF) at Lawrence Livermore National Laboratory have achieved success in producing more energy than the laser pulse that was used to heat the fuel. The NIF focused on inertial confinement fusion, which uses a system of lasers to heat up fuel pellets producing a plasma—a cloud of charged ions. The fuel pellets contain “heavy” versions of hydrogen—deuterium and tritium—that are easier to fuse and produce more energy. However, the fuel pellets need to be heated and pressurised to conditions found at the centre of the Sun, which is a natural fusion reactor. Once these conditions are achieved, fusion reactions release several particles, including alpha particles, which interact with the surrounding plasma and heat it up further. The heated plasma then releases more alpha particles and so on, in a self-sustaining reaction—a process referred to as ignition.
New technique found to simplify qubits manufacture
Scientists from Japan have discovered how a superconducting material, niobium nitride, can be added as a flat, crystalline layer to a nitride-semiconductor substrate. This technique could make it simple to manufacture quantum qubits that can be used with conventional computer devices. Niobium nitride can become superconducting at temperatures 16°C above absolute zero. Thus, it can be utilised to create a superconducting qubit when arranged in a structure called a Josephson junction. Researchers investigated the impact of temperature on the crystal structures and electrical properties of NbNx thin films grown on AlN template substrates. They showed that the spacing of atoms in the two materials was compatible enough to produce flat layers. This interface between the AlN substrate, which has a wide bandgap, and NbNx is essential for future quantum devices, such as Josephson junctions. Superconducting layers that are only a few nanometers thick and have high crystallinity can be used as detectors of single photons or electrons.
New technique found to improve free-space optical communication
New signal-processing algorithms have been shown to mitigate the impact of turbulence in free-space optical experiments, potentially bringing “free space” internet a step closer to reality. A team of researchers from Aston University’s Aston Institute of Photonic Technologies and Glasgow University used commercially available photonic lanterns, a commercial transponder, and a spatial light modulator to emulate turbulence. They simultaneously transmitted multiple data signals using different spatially shaped beams of light using a so-called photonic lantern. Turbulence changes the shape of the beams, often losing the signal if only a single simple shape is transmitted and detected, but by detecting light with these shapes using a second lantern, more of the light is collected at the receiver, and the original data can be unscrambled. This can greatly reduce the impact of the atmosphere on the quality of the data received in a technique known as Multiple-Input MultipleOutput (MIMO) digital signal processing. This project aims to provide the internet performance of a pure fibre connection without the need to install cables.
New optical approach can enable much faster communication
A new all-optical approach for driving multiple highly dense nano laser arrays has been developed by researchers in Korea. This method can enable chip based optical communication links that process and move data much faster than current electronic based devices. Researchers have demonstrated that densely integrated nano laser arrays—in which the lasers are just 18 microns apart—can be fully driven and programmed with light from a single optical fibre. Researchers used electrodes with a unique optical driver that creates programmable patterns of light via interference. This light travels through an optical fibre onto which nano lasers are printed. To demonstrate this approach, the researchers used a high-resolution transfer-printing technique to fabricate multiple photonic crystal nano lasers spaced 18 microns apart. These arrays were applied onto the surface of a 2-micron-diameter optical microfibre. The interference pattern could also be modified by adjusting the driving beam’s polarisation and pulse width. The design allowed multiple nano laser arrays to be driven using light traveling through a single fibre.
ML model to predict earthquakes and pandemics developed
Researchers at Brown University and MIT have developed a machine-learning model that can predict an upcoming natural calamity like earthquakes and pandemics. They combined statistical algorithms—which need less data to make accurate, efficient predictions—with a powerful machine learning technique and trained the model to predict scenarios, probabilities, and sometimes even the timeline of rare events despite the lack of historical record on them. Researchers applied the pinpointing parameter and different ranges of probabilities for dangerous spikes during a pandemic, finding and predicting rogue waves, and estimating when a ship will crack in half due to stress. They found they could discover and quantify when rogue waves will form by looking at probable wave conditions that nonlinearly interact over time, leading to waves sometimes three times their original size.
Flaws in lithium-ion batteries being studied to improve them
An international group of experts, led by researchers from Texas A&M University, have combined powerful imaging techniques and large data sets to better understand why lithium-ion batteries fail and how they can be improved. Researchers were able to visualise the flaws and map defects that occur within the batteries using the SM beamline at the Canadian Light Source (CLS)—located at the University of Saskatchewan (UofS). The team is interested in identifying in real time the flaws that occur, which would allow them to measure the complex relationship between the battery’s materials, its shape, and the chemical reactions that occur within. Reaching this goal would allow researchers to proactively address design failures. More efficient batteries can help reduce waste and help us transition to a greener grid.
Objects steered with the help of ultrasonic waves
Researchers at University of Minnesota have developed a way to freely manipulate objects with the help of ultrasound waves. Their method can move larger objects using the principles of metamaterial physics. Metamaterials are artificially designed materials that can interact with waves like light and sound. By placing a metamaterial pattern on the surface of an object, the researchers were able to use sound to steer it in a certain direction without physically touching it. When these tiny patterns are placed on the surface of the objects, one can basically reflect the sound in any direction. And in doing that, the user can control the acoustic force that is exerted on an object. Using this technique, the researchers can not only move an object forward but also pull it toward a source—not too dissimilar from the “force” described in Star Wars.
A new system devised to teach deep learning more efficiently
Researchers at Texas A&M University, Rain Neuromorphics, and Sandia National Laboratories have devised a new system for training deep learning models more efficiently and on a larger scale. This approach could reduce the carbon footprint and financial costs associated with the training of AI models, thus making their large-scale implementation easier and more sustainable. To do this, they had to over- come two key limitations of current AI training practices. The first challenge was the use of inefficient hardware systems based on GPUs, whereas the second entails the use of ineffective and math-heavy software tools, specifically utilising the so-called backpropagation algorithm. Researchers developed a new co-optimised learning algorithm that exploits the hardware parallelism of memristor crossbars. This algorithm, inspired by the differences in neuronal activity observed in neuroscience studies, is tolerant to errors and replicates the brain’s ability to learn even from sparse, poorly defined, and “noisy” information.
Unique nano chips formed using silicon carbide chips
Researchers at the Georgia Institute of Technology have created a modified form of epi graphene on a silicon carbide crystal substrate to create a new nanoelectronics platform. They produced unique silicon carbide chips from electronics-grade silicon carbide crystals. They used electron beam
lithography, a method commonly used in microelectronics, to carve the graphene nanostructures and weld their edges to the silicon carbide chips. This process mechanically stabilises and seals the graphene’s edges, which would otherwise react with oxygen and other gases that might interfere with the motion of the charges along the edge. Results showed
that the graphene edge states were similar to photons in an optical fibre that can travel over large distances without scattering. They found that the charges traveled for tens of thousands of nanometres along the edge before scattering.
Eco-friendly wirelessly powered IoT sensors on the way
KAUST alumni Kalaivanan Loganathan, with Thomas Anthopoulos and coworkers, assessed the viability of various large-area electronic technologies and their potential to deliver eco-friendly, wirelessly powered IoT sensors. He has developed a range of RF electronic components, including metal-oxide and organic polymer based semiconductor devices known as Schottky diodes. “These devices are crucial components in wireless energy harvesters and ultimately dictate the performance and cost of the sensor nodes,” Loganathan says. Key contributions from the KAUST team include scalable methods for manufacturing RF diodes to harvest energy reaching the 5G/6G frequency range. Such technologies provide the needed building blocks toward a more sustainable way to power the billions of sensor nodes in the near future. The team is currently working on the monolithic integration of these low-power devices with antenna and sensors to showcase their true potential.
Spray-on sensory system makes hands smarter
Researchers from Stanford University have developed a spray-on sensory system which consists of printed, biocompatible nanomesh directly connected with a wireless Bluetooth module and further trained through meta-learning. The device can identify objects by touch alone or allow users to communicate by hand gestures with apps in immersive environments. The spray-on approach enables the device to function without any need of a substrate. This decision eliminated unwanted motion artifacts and allowed them to use a single trace of conductive mesh to generate multi-joint information of the fingers. The device functions on the basis of a machine learning algorithm. Computers monitor the changing patterns in conductivity and map those changes to specific physical tasks and gestures. To meet this intense computational challenge and the need to compute vast amounts of data, the Stanford team has developed a learning scheme that is far more computationally efficient.
Earbuds shown to work as hearing aids much more economically
Recent studies have demonstrated the effectiveness of earbuds in case of hearing loss. According to these studies, commercial earbuds can be adapted as hearing aids in amplifying sound for individuals with hearing loss. One such study was conducted by a team of researchers led by Yen-fu Cheng, an otolaryngologist at Taipei Veterans General Hospital. The team compared AirPods 2 and AirPods Pro—the model with a noise canceling feature—with a set of premium hearing aids and a basic pair of hearing aids. Notably, AirPods Pro met four out of five technology standards for hearing aids. It was also observed that both AirPods models were far cheaper than the basic as well as the premium hearing aids.
Wireless communication at terahertz speed shown over large distances
Researchers have proved that high-speed, high-bandwidth wireless communication at the terahertz frequency is possible across long distances. Their work shows that there is a path forward for mass wireless communication, one that could shrink the digital divide felt by rural communities outside high-speed optical fibre networks. Researchers were able to form a 2km link, the longest terahertz connection ever established on Earth.
The team directly fed the information to the signal source. The results showed that the new system was hitting frequencies and bandwidths that eclipsed 5G networks by more than “two orders of magnitude.” The impact of this kind of high-speed, high bandwidth connection would be monumental, providing higher data rates and more connectivity even for rural communities.
Solar reactor converts plastic and greenhouse gasses to fuel
Researchers at the University of Cambridge have developed a system that can transform plastic waste and greenhouse gasses into sustainable fuels and other valuable products—using just the energy from the Sun. It can convert two waste streams into two chemical products at the same time—the first time this has been achieved in a solar-powered reactor. The reactor converts the carbon dioxide and plastics into different products which can be utilised in a variety of industries. Researchers were able to convert carbon dioxide into syngas, a key building block for sustainable liquid fuels, and plastic bottles were converted into glycolic acid, which is widely used in the cosmetics industry. They also designed different catalysts, which were integrated into the light absorber. Changing the catalyst enabled researchers to then change the end product.