Lifeboat Foundation

The maker of a defunct cloud photo storage app that pivoted to selling facial recognition services has been ordered to delete user data and any algorithms trained on it, under the terms of an FTC settlement.

The regulator investigated complaints the Ever app — which gained earlier notoriety for using dark patterns to spam users’ contacts — had applied facial recognition to users’ photographs without properly informing them what it was doing with their selfies.

Under the proposed settlement, Ever must delete photos and videos of users who deactivated their accounts and also delete all face embeddings (i.e. data related to facial features which can be used for facial recognition purposes) that it derived from photos of users who did not give express consent to such a use.

Two researchers at Duke University have recently devised a useful approach to examine how essential certain variables are for increasing the reliability/accuracy of predictive models. Their paper, published in Nature Machine Intelligence, could ultimately aid the development of more reliable and better performing machine-learning algorithms for a variety of applications.

“Most people pick a predictive machine-learning technique and examine which variables are important or relevant to its predictions afterwards,” Jiayun Dong, one of the researchers who carried out the study, told TechXplore. “What if there were two models that had similar performance but used wildly different variables? If that was the case, an analyst could make a mistake and think that one variable is important, when in fact, there is a different, equally good model for which a totally different set of variables is important.”

Dong and his colleague Cynthia Rudin introduced a method that researchers can use to examine the importance of variables for a variety of almost-optimal predictive models. This approach, which they refer to as “variable importance clouds,” could be used to gain a better understanding of machine-learning models before selecting the most promising to complete a given task.

The French theoretical physicist Franck Laloë presents a modification of Schrödinger’s famous equation that ensures that all measured states are unique, helping to solve the problem that is neatly encompassed in the Schördinger’s cat paradox.

The paradox of Schrödinger’s cat – the feline that is, famously, both alive and dead until its box is opened – is the most widely known example of a recurrent problem in quantum mechanics: its dynamics seems to predict that macroscopic objects (like cats) can, sometimes, exist simultaneously in more than one completely distinct state. Many physicists have tried to solve this paradox over the years, but no approach has been universally accepted. Now, however, theoretical physicist Franck Laloë from Laboratoire Kastler Brossel (ENS-Université PSL) in Paris has proposed a new interpretation that could explain many features of the paradox. He sets out a model of this possible theory in a new paper in EPJ D.

One approach to solving this problem involves adding a small, random extra term to the Schrödinger equation, which allows the quantum state vector to ‘collapse’, ensuring that – as is observed in the macroscopic universe – the outcome of each measurement is unique. Laloë’s theory combines this interpretation with another from de Broglie and Bohm and relates the origins of the quantum collapse to the universal gravitational field. This approach can be applied equally to all objects, quantum and macroscopic: that is, to cats as much as to atoms.

Recently, there has been a reemergence of interest in optical computing platforms for artificial intelligence-related applications. Optics is ideally suited for realizing neural network models because of the high speed, large bandwidth and high interconnectivity of optical information processing. Introduced by UCLA researchers, Diffractive Deep Neural Networks (D²NNs) constitute such an optical computing framework, comprising successive transmissive and/or reflective diffractive surfaces that can process input information through light-matter interaction. These surfaces are designed using standard deep learning techniques in a computer, which are then fabricated and assembled to build a physical optical network. Through experiments performed at terahertz wavelengths, the capability of D²NNs in classifying objects all-optically was demonstrated. In addition to object classification, the success of D²NNs in performing miscellaneous optical design and computation tasks, including e.g., spectral filtering, spectral information encoding, and optical pulse shaping have also been demonstrated.

In their latest paper published in Light: Science & Applications, UCLA team reports a leapfrog advance in D²NN-based image classification accuracy through ensemble learning. The key ingredient behind the success of their approach can be intuitively understood through the experiment of Sir Francis Galton (1822–1911), an English philosopher and statistician, who, while visiting a livestock fair, asked the participants to guess the weight of an ox. None of the hundreds of participants succeeded in guessing the weight. But to his astonishment, Galton found that the median of all the guesses came quite close—1207 pounds, and was accurate within 1% of the true weight of 1198 pounds. This experiment reveals the power of combining many predictions in order to obtain a much more accurate prediction. Ensemble learning manifests this idea in machine learning, where an improved predictive performance is attained by combining multiple models.

In their scheme, UCLA researchers reported an ensemble formed by multiple D²NNs operating in parallel, each of which is individually trained and diversified by optically filtering their inputs using a variety of filters. 1252 D²NNs, uniquely designed in this manner, formed the initial pool of networks, which was then pruned using an iterative pruning algorithm, so that the resulting physical ensemble is not prohibitively large. The final prediction comes from a weighted average of the decisions from all the constituent D²NNs in an ensemble. The researchers evaluated the performance of the resulting D²NN ensembles on CIFAR-10 image dataset, which contains 60000 natural images categorized in 10 classes and is an extensively used dataset for benchmarking various machine learning algorithms. Simulations of their designed ensemble systems revealed that diffractive optical networks can significantly benefit from the ‘wisdom of the crowd’.

Capturing energy from the Sun with solar panels is only half the story – that energy needs to be stored somewhere for later use. In the case of flow batteries, storage is relegated to vats of liquid. Now, an international team led by University of Wisconsin-Madison scientists has created a new version of these solar flow batteries that’s efficient and long-lasting.

To make the new device, the team combined several existing technologies. It’s a silicon/perovskite tandem solar cell, paired with a redox flow battery, which the team says will allow people to harvest and store renewable energy in one device. Not only is it efficient, but it should be inexpensive and simple enough to scale up for home use.

The energy-harvesting part of the equation combines the long-time industry-leading material – silicon – with a promising young upstart called perovskite. These tandem solar cells have proved better than either material alone, since the two materials capture different wavelengths of light.

The cable, called hollow-core fiber, is a next-generation version of the fiber-optic cable used to deliver broadband internet to homes and businesses. Made of glass, such cables carry data encoded as beams of light. But instead of being solid, hollow-core fiber is empty inside, with dozens of parallel, air-filled channels narrower than a human hair.

Because light travels nearly 50% faster through air than glass, it takes about one-third less time to send data through hollow-core fiber than through the same length of standard fiber.

The difference is often just a minuscule fraction of a second. But in high-frequency trading, that can make the difference between profits and losses. HFT firms use sophisticated algorithms and ultrafast data networks to execute rapid-fire trades in stocks, options and futures. Many are secretive about their trading strategies and technology.

Smartwatches and other battery-powered electronics would be even smarter if they could run AI algorithms. But efforts to build AI-capable chips for mobile devices have so far hit a wall—the so-called “memory wall” that separates data processing and memory chips that must work together to meet the massive and continually growing computational demands imposed by AI.

“Transactions between processors and memory can consume 95 percent of the energy needed to do machine learning and AI, and that severely limits battery life,” said computer scientist Subhasish Mitra, senior author of a new study published in Nature Electronics.

Now, a team that includes Stanford computer scientist Mary Wootters and electrical engineer H.-S. Philip Wong has designed a system that can run AI tasks faster, and with less energy, by harnessing eight hybrid chips, each with its own data processor built right next to its own memory storage.

A new deep-learning algorithm could provide advanced notice when systems — from satellites to data centers — are falling out of whack.

When you’re responsible for a multimillion-dollar satellite hurtling through space at thousands of miles per hour, you want to be sure it’s running smoothly. And time series can help.

A time series is simply a record of a measurement taken repeatedly over time. It can keep track of a system’s long-term trends and short-term blips. Examples include the infamous Covid-19 curve of new daily cases and the Keeling curve that has tracked atmospheric carbon dioxide concentrations since 1958. In the age of big data, “time series are collected all over the place, from satellites to turbines,” says Kalyan Veeramachaneni. “All that machinery has sensors that collect these time series about how they’re functioning.”

(Inside Science) — It took years of painstaking work for Carlos Souza and his colleagues to map out every road in the Brazilian Amazon biome. Official maps of the 4.2 million-square-kilometer region only show roads built by federal and local governments. But by carefully tracing lines on satellite images, the researchers concluded in 2016 that the true length of all the roads combined was nearly 13 times higher.

“When we don’t have a good understanding of how much roadless areas we have on the landscape, we probably will misguide any conservation plans for that territory,” said Souza, a geographer at a Brazil-based environmental nonprofit organization called Imazon.

Now, Imazon researchers have built an artificial intelligence algorithm to find such roads automatically. Currently, the algorithm is reaching about 70% accuracy, which rises to 87%-90% with some additional automated processing, said Souza. Analysts then confirm potential roads by examining the satellite images. Souza presented the research last month at a virtual meeting of the American Geophysical Union.