Google users contribute more than 20 million pieces of information on Maps every day – that's more than 200 contributions every second. The uncertainty of traffic can crash the algorithms predicting the best ETA. There is also a chance of new roads and buildings being built all the time. Though Google Maps gets its ETA right most of the time, there is still room for improvement. Researchers at Alphabet-owned DeepMind have partnered with the Google Maps team to improve the accuracy of the real-time ETAs by up to 50% in places like Berlin, Jakarta, São Paulo, Sydney, Tokyo, and Washington D.C.

In October 2019, Google announced that it would process search queries with the BERT model that its researchers have developed. This model can grasp difficult nuances of language: in the search 2019 brazil traveler to the USA need a visa, it is understood that the traveler is Brazilian, and the destination is the USA. From now on, Google says, this search returns the page of the U.S. embassy in Brazil, and no longer shows a page about U.S. citizens traveling to Brazil. Remarkably, most of the attention mechanism at the core of many transformer models like BERT relies on just a few basic vector operations. Let's see how it works.

SCI COMMUN### Astronomy Talk about a sharper image: A recently constructed imaging sensor array (above) that will be used when the Vera C. Rubin Observatory in Chile opens in 2021 has captured a world-record 3200 megapixels in a single shot. It recorded a variety of objects, including a Romanesco broccoli, at that resolution, which is detailed enough to show a golf ball clearly from 24 kilometers away. The sensor array's focal plane is more than 60 centimeters wide, much larger than the 3.5-centimeter sensors on high-end consumer digital cameras, says the SLAC National Accelerator Laboratory, which built the array. When the telescope, funded by the U.S. National Science Foundation, begins operating next year, it will image the entire southern sky every few nights for 10 years, cataloguing billions of galaxies each time. The surveys will shed light on mysterious dark energy and dark matter, which make up most of the universe's mass. With its repeat coverage, the telescope will make the equivalent of an astronomical movie in order to discover objects that suddenly appear, move, or go bang. ### Biomedicine Corticosteroids given orally or intravenously should be the standard therapy for people with “severe and critical” COVID-19, the World Health Organization (WHO) said in new guidelines issued last week—but they should not be given to patients with mild cases. In June, a large U.K. trial named Recovery first showed that the steroid dexamethasone cut deaths among ventilated COVID-19 patients by 35% after 28 days of treatment. That result was confirmed by a WHO-sponsored metaanalysis published in JAMA on 2 September that included Recovery and six other studies testing dexamethasone, as well as two other corticosteroids—hydrocortisone and methylprednisolone. Many countries, including the United States, had already included corticosteroids in their national treatment guidelines. But WHO's recommendations will be important as a signal to low- and middle-income countries, says Martin Landray, one of Recovery's principal investigators. ### Public health COVID-19 virus particles drifting through a Chinese apartment building's plumbing may have infected some residents, a study has found, raising fears of yet another way that the disease could spread. The case echoes a 2003 outbreak of severe acute respiratory syndrome (SARS) that spread through the pipes of a Hong Kong apartment building. Such transmission is difficult to prove. But scientists suspect that aerosolized coronavirus may have spread from the bathroom of a Guangzhou family of five through a floor drain and into the building's wastewater pipes. Two middle-aged couples living in apartments above the family later contracted COVID-19. The study appeared last week in Annals of Internal Medicine . ### Conservation A plan to reforest a cross-continental strip of Africa to hold back expansion of the Sahara Desert and the semi-arid Sahel has made little progress—even though the project is halfway toward its planned completion date in 2030, a report says. Participating countries have planted only 4 million hectares of trees and other vegetation for the Great Green Wall, well short of the 100 million planned to stretch 7000 kilometers from Senegal to Djibouti, says the report by the Climatekos consulting firm, presented on 7 September at a meeting of the countries' ministers. Supporters predicted the project would also create jobs and capture carbon dioxide. Scientists have said creating grasslands may be more effective than planting trees to resist desertification, The Guardian reported. ### Philanthropy Rice University last week received a $100 million gift for materials science. It is the largest to date in that discipline recorded in a database of gifts for engineering maintained by The Chronicle of Philanthropy . The funding will be used to pair materials science with artificial intelligence to advance the design and manufacturing of new materials, for applications that include sustainable water systems, energy, and telecommunications. The donor was the Robert A. Welch Foundation, which supports chemistry research in Texas. ### Conservation Scientists hailed a move last week by the European Union to ban the use of lead ammunition near wetlands and waterways. The European Chemicals Agency has estimated that as many as 1.5 million aquatic birds die annually from lead poisoning because they swallow some of the 5000 tons of lead shot that land in European wetlands each year. Its persistence in the environment is also considered a human health hazard. The EU Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) committee approved the ban after years of controversy. The German delegation, which had abstained in a July vote on the issue, changed its stance to support the measure after a letter from 75 scientists and petitions signed by more than 50,000 people called for it to do so. The European Commission and the European Parliament are expected to formally approve the ban, allowing it to go into effect in 2022. REACH may debate a complete ban on lead ammunition and fishing weights later this year. ### Chemical weapons Alexei Navalny, a Russian opposition politician, was poisoned with a nerve agent “identified unequivocally in tests” as a Novichok, an exotic Sovietera chemical weapon, German Chancellor Angela Merkel said on 2 September. Navalny fell ill on 20 August after drinking a cup of tea at a Siberian airport. He was flown to Berlin and this week emerged from a coma. German military scientists at the Bundeswehr Institute of Pharmacology and Toxicology in Munich haven't released details of their tests, but they had clear targets to hunt for: Like other nerve agents, Novichoks bind to the enzymes acetylcholinesterase and butyrylcholinesterase, creating a telltale conjugate compound. Novichok agents came to wide public notice in 2018 after one was used in an assassination attempt against former Russian spy Sergei Skripal in the United Kingdom. The attack prompted nations to push for a crackdown on Novichok agents, and last year they were added to the list of toxic chemicals regulated under the Chemical Weapons Convention. ### COVID-19 In one of the largest surveys of Americans since COVID-19 lockdowns began, a majority reported having some symptoms of depression, up from one-quarter in a prepandemic survey. The prevalence of symptoms graded as moderate to severe tripled, to 27.8% of respondents. A research team compared results from two surveys used to screen for depression: one administered to more than 5000 people in 2017 and 2018 by the U.S. Centers for Disease Control and Prevention, the other given to 1400 people in early April by NORC at the University of Chicago. Prevalence of depression symptoms rose in all demographic groups and especially among individuals facing financial problems, job loss, or family deaths. The increases in self-reported symptoms are larger than those recorded in previous surveys after large-scale traumatic events in other countries, including outbreaks of the severe acute respiratory syndrome, H1N1, and Ebola, the authors write in the 2 September issue of JAMA Network Open . ### A U.S. vaccine leader's vow: Politics stays out “I would immediately resign if there is undue interference in this process.” So said Moncef Slaoui, scientific director of Operation Warp Speed, the U.S. effort to quickly develop a vaccine for COVID-19, in an interview with Science . To date, Warp Speed has invested more than $10 billion in eight vaccine candidates. Three are now in large-scale efficacy trials, and interim reviews of their data by independent safety and monitoring boards could reveal evidence of protection as early as October. Slaoui, an immunologist who formerly headed vaccine development at GlaxoSmithKline, answered questions from Science last week about how Warp Speed operates and addressed concerns that political pressure before the 3 November U.S. presidential election may lead to an emergency use authorization of a COVID-19 vaccine before it is proven safe and effective. (On 8 September, nine companies developing vaccines for the pandemic coronavirus pledged not to seek a premature authorization.) “It needs to be absolutely shielded from the politics,” Slaoui says. “Trust me, there will be no [authorization request] filed if it's not right. … The science is what is going to guide us. … And at the end of the day, the facts and the data will be made available to everyone who wants to look at them and will be transparent.” Slaoui defended Warp Speed's decision to not consider vaccines made of whole, inactivated viruses, a time-tested approach. China has three such vaccines in efficacy trials, but he worries they could cause serious side effects in people who receive them. Slaoui also said if it had been his choice, the United States would have participated in COVAX, a mechanism for countries to invest collectively in vaccines and share them; the Trump administration declined to join. The full interview—one of Slaoui's most detailed since taking the job in May—is at .

If you want to create a world-class recommendation system, follow this recipe from a global team of experts: Blend a big helping of GPU-accelerated AI with a dash of old-fashioned cleverness. The proof was in the pudding for a team from NVIDIA that won this year's ACM RecSys Challenge. The competition is a highlight of an annual gathering of more than 500 experts who present the latest research in recommendation systems, the engines that deliver personalized suggestions for everything from restaurants to real estate. At the Sept. 22-26 online event, the team will describe its dish, already available as open source code. They're also sharing lessons learned with colleagues who build NVIDIA products like RAPIDS and Merlin, so customers can enjoy the fruits of their labor.

The companies claim BioAssist and LaborAssist can deliver a better understanding of a patient's birthing progress by automatically taking metrics like fetal angle of progression during labor. According to the World Health Organization, about 295,000 women died during pregnancy and following childbirth in 2017. Research from the Perinatal Institute shows that tracking fetal growth can help prevent stillbirths, as it allows physicians to recognize growth restrictions. But there's little standardization around fetal measurement because doctors often disagree on where to start. Intel and Samsung Medison say BioAssist and LaborAssist perform measurements in as little as 85 microseconds, courtesy of Intel's machine learning toolkit OpenVino.

Google Maps takes assistance from Deep Minds, a London-based AI lab owned by Google's parent company Alphabet to get accurate results for users. Google, in a blog post, explains how the AI helps it collect data using neural networks to predict future traffic. This data includes live traffic information collected anonymously from Android devices, historical traffic data, information like speed limits and construction sites from local governments, and also factors like the quality, size, and direction of any given road, The Verge noted. "We then combine this database of historical traffic patterns with live traffic conditions, using machine learning to generate predictions based on both sets of data," Johann Lau, product Manager Google Maps, wrote in the blog post. Other than the AI, Google relies on traffic data from authorities of local governments and real-time feedback from users.

Chess has a reputation for cold logic, but Vladimir Kramnik loves the game for its beauty. "It's a kind of creation," he says. His passion for the artistry of minds clashing over the board, trading complex but elegant provocations and counters, helped him dethrone Garry Kasparov in 2000 and spend several years as world champion. Yet Kramnik, who retired from competitive chess last year, also believes his beloved game has grown less creative. He partly blames computers, whose soulless calculations have produced a vast library of openings and defenses that top-flight players know by rote.

If you are one of those who start their journey by setting the route on Google Maps, Google brings you a better version of Google Maps. DeepMind, the London-based AI lab owned by Google's parent company Alphabet, has developed a new Machine Learning algorithm to combine live traffic conditions with historical traffic patterns for more accurate Estimated Time of Arrival(ETA). This data includes live traffic information collected anonymously from Android devices, historical traffic data, information like speed limits, and factors like quality, size, and direction. DeepMind has reduced the percentage of prediction inaccuracies drastically. In one of the test models, the prediction errors dropped by less than 51%.

Chess has a reputation for cold logic, but Vladimir Kramnik loves the game for its beauty. "It's a kind of creation," he says. His passion for the artistry of minds clashing over the board, trading complex but elegant provocations and counters, helped him dethrone Garry Kasparov in 2000 and spend several years as world champion. Yet Kramnik, who retired from competitive chess last year, also believes his beloved game has grown less creative. He partly blames computers, whose soulless calculations have produced a vast library of openings and defenses that top flight players know by rote.

Risk is 6 player game with significant randomness and a large game-tree complexity which poses a challenge to creating an agent to play the game effectively. Previous AIs focus on creating high-level handcrafted features determine agent decision making. In this project, I create D.A.D, A Risk agent using temporal difference reinforcement learning to train a Deep Neural Network including a Graph Convolutional Network to evaluate player positions. This is used in a game-tree to select optimal moves. This allows minimal handcrafting of knowledge into the AI, assuring input features are as low-level as possible to allow the network to extract useful and sophisticated features itself, even with the network starting from a random initialisation. I also tackle the issue of non-determinism in Risk by introducing a new method of interpreting attack moves necessary for the search. The result is an AI which wins 35% of the time versus 5 of best inbuilt AIs in Lux Delux, a Risk variant.