This week’s list of data news highlights covers July 21-27, 2018, and includes articles about genetic clues that could indicate how long people stay in school and a database tracking human rights abuses in Syria.
The U.S. Defense Advanced Projects Research Agency (DARPA) has a new research initiative called the Artificial Intelligence Exploration (AIE) program devoted to advancing “third wave” AI theory and applications, which focus on allowing machines to contextually adapt to new situations. “First wave” AI emphasized rule-based problem solving and “second wave” AI emphasized statistical learning, however these approaches were limited in that they were inflexible and only narrowly effective.
Researchers at the University of Virginia working on the largest-ever study on the relationship between genes and human cognition has identified more than 1,000 genetic markers that seem to influence how long a person will stay in school. The researchers analyzed DNA from 1.1 million people and found that people with the lowest genetic scores, as indicated by the presence of these markers, had only a 10 percent chance of having graduated college, while those with the highest scores had a 50 percent chance of doing so. The researchers used data from the UK Biobank as well as data provided by 23andMe.
Thailand’s state-owned telecom company CAT Telecom has announced plans to deploy a low-power wide-area network (LoRaWAN) across the country to support smart city applications. The initiative is part of the government’s Thailand 4.0 plan to modernize Thailand’s economy and foster innovation. CAT Telecom will deploy the network initially in Phuket and Bangkok for smart city projects before launching it in other areas.
Researchers at Los Alamos National Lab and Carnegie Mellon University have used the Trinity supercomputer, one of the top 10 fastest in the world, to successfully simulate a trillion high-speed particles in a record-breaking speed of two minutes. The researchers were studying a phenomenon called the Fermi acceleration, which occurs when particles in supernovae and solar flares rapidly gain speed, and wanted to know which particles out of a trillion would reach the highest speeds. Rather than take snapshots of one trillion particles as they accelerated to review and identify which ones sped up the fastest, as in a normal simulation, the researchers instead took snapshots of every individual particle, creating more files overall but each with less data. This allowed the researchers to identify the fastest few thousand particles in just two minutes, thousands of times faster than it would have taken to analyze the simulation using traditional approaches.
Google has developed a specialized miniature microchip it calls the Edge Tensor Processing Unit (TPU) that can run on-device AI for enterprise applications. On-device AI can allow for applications such as automating quality control checks in factories without the need to transmit data to a central computer for processing, which can be slower and subject to greater downtime.
The Violations Documentation Center (VDC), a Copenhagen-based non-governmental organization, has created a database tracking incidents of human rights abuses spanning the eight-year long Syrian civil war. The database contains 600 terabytes of images, witness testimonies, and data about cause of death, location, military involvement, and weapons used. Of the 500,000 people estimated to have been in the war, VDC has data about 188,957 of those deaths, mostly civilians, of which it estimates 77 percent were a result of the Assad regime violating humanitarian law.
Researchers at the Massachusetts Institute of Technology have developed a machine learning system that can analyze subtle facial expressions to perceive a person’s mood. Affective computing, which focuses on developing AI that can analyze factors such as facial expression of voice tone to gauge emotion and respond accordingly has struggled with interpreting the variations in how different people express the same emotions. The researchers trained their system on thousands of images of faces with subtle variations in their expressions allowing it to interpret the emotions of new faces with similar efficacy and more accurately than existing systems.
The U.S. National Institutes of Health (NIH) have launched the Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability (STRIDES) Initiative to facilitate partnerships that can make advanced cloud-based analytics technologies more accessible to researchers. STRIDES will focus on making high-value datasets available through the cloud and developing cloud-based machine learning tools for medical research. NIH has selected Google Cloud as its first industry partner.
Researchers at DeepMind have created an AI system called ToMnet (Theory of Mind net) that uses three neural networks to observe other AI systems and learn about how they work. ToMnet’s first neural network analyzes the past actions of other AIs, while the second attempts to identify their “beliefs,” and the third combines the outputs of the first two networks and attempts to predict an AI systems future actions. ToMnet and similar theory of mind research could eventually allow developers to explain how complex “black box” AI systems operate.
23andMe has partnered with GlaxoSmithKline to share anonymized genetic data from 5 million customers with the pharmaceutical company to facilitate the development of new drugs. 23andMe has previously partnered with GlaxoSmithKline to help identify people with Parkinson’s disease caused by a rare genetic condition, which allowed GlaxoSmithKline to recruit 250 patients for clinical trials.
Image: Максим Улитин.