"Britons on way to France risk quarantine" is the front page headline in the Times, as it reports that Whitehall officials have placed the country on a list of destinations to be closely monitored. A senior aviation source is quoted saying France is "bubbling" with cases and that travellers should only book trips which can be re-arranged at 24 hours' notice. The Daily Telegraph also reports the close monitoring of France as cases there overtake the numbers for Portugal, which has reduced its infection rate. The paper says about 450,000 Britons are currently holidaying in France, a scale which would make any new restrictions a logistical nightmare. The Guardian leads with an exclusive warning from doctors' leaders that shutting down non-Covid NHS services to deal with any second wave will leave thousands of patients unacceptably "stranded", risking more deaths.

Ideally Open-Domain Question Answering models should exhibit a number of competencies, ranging from simply memorizing questions seen at training time, to answering novel question formulations with answers seen during training, to generalizing to completely novel questions with novel answers. However, single aggregated test set scores do not show the full picture of what capabilities models truly have. In this work, we perform a detailed study of the test sets of three popular open-domain benchmark datasets with respect to these competencies. We find that 60-70% of test-time answers are also present somewhere in the training sets. We also find that 30% of test-set questions have a near-duplicate paraphrase in their corresponding training sets. Using these findings, we evaluate a variety of popular open-domain models to obtain greater insight into what extent they can actually generalize, and what drives their overall performance. We find that all models perform dramatically worse on questions that cannot be memorized from training sets, with a mean absolute performance difference of 63% between repeated and non-repeated data. Finally we show that simple nearest-neighbor models out-perform a BART closed-book QA model, further highlighting the role that training set memorization plays in these benchmarks

This paper describes our winning solution for the ECML-PKDD ChAT Discovery Challenge 2020. We show that whether or not a Twitch user has subscribed to a channel can be well predicted by modeling user activity with boosting trees. We introduce the connection between target-encodings and boosting trees in the context of high cardinality categoricals and find that modeling user activity is more powerful then direct modeling of content when encoded properly and combined with a gradient boosting optimization approach.

All famous machine learning algorithms that comprise both supervised and semi-supervised learning work well only under a common assumption: the training and test data follow the same distribution. When the distribution changes, most statistical models must be reconstructed from newly collected data, which for some applications can be costly or impossible to obtain. Therefore, it has become necessary to develop approaches that reduce the need and the effort to obtain new labeled samples by exploiting data that are available in related areas, and using these further across similar fields. This has given rise to a new machine learning framework known as transfer learning: a learning setting inspired by the capability of a human being to extrapolate knowledge across tasks to learn more efficiently. Despite a large amount of different transfer learning scenarios, the main objective of this survey is to provide an overview of the state-of-the-art theoretical results in a specific, and arguably the most popular, sub-field of transfer learning, called domain adaptation. In this sub-field, the data distribution is assumed to change across the training and the test data, while the learning task remains the same. We provide a first up-to-date description of existing results related to domain adaptation problem that cover learning bounds based on different statistical learning frameworks.

According to the National Academies, a weekly forecast of velocity, vertical structure, and duration of the Loop Current (LC) and its eddies is critical for understanding the oceanography and ecosystem, and for mitigating outcomes of anthropogenic and natural disasters in the Gulf of Mexico (GoM). However, this forecast is a challenging problem since the LC behaviour is dominated by long-range spatial connections across multiple timescales. In this paper, we extend spatiotemporal predictive learning, showing its effectiveness beyond video prediction, to a 4D model, i.e., a novel Physics-informed Tensor-train ConvLSTM (PITT-ConvLSTM) for temporal sequences of 3D geospatial data forecasting. Specifically, we propose 1) a novel 4D higher-order recurrent neural network with empirical orthogonal function analysis to capture the hidden uncorrelated patterns of each hierarchy, 2) a convolutional tensor-train decomposition to capture higher-order space-time correlations, and 3) to incorporate prior physic knowledge that is provided from domain experts by informing the learning in latent space. The advantage of our proposed method is clear: constrained by physical laws, it simultaneously learns good representations for frame dependencies (both short-term and long-term high-level dependency) and inter-hierarchical relations within each time frame. Experiments on geospatial data collected from the GoM demonstrate that PITT-ConvLSTM outperforms the state-of-the-art methods in forecasting the volumetric velocity of the LC and its eddies for a period of over one week.

SCI COMMUN### Conservation A company seeking to build a controversial gold and copper mine in Alaska won a major victory on 24 July when the U.S. Army Corps of Engineers issued an environmental analysis saying the mine wouldn't endanger the world's most productive sockeye salmon fishery. The decision clears the way for the Corps to issue permits needed by promoters of the Pebble Mine, located at the headwaters of two major watersheds that form part of the Bristol Bay salmon runs, just north of the Aleutian Islands. Environmental and Native Alaskan groups and some salmon scientists blasted the new study, saying it understated risks by focusing on the mine's small, initial footprint over 20 years of mining rather than its potential impacts if it expands to become one of the world's largest gold and copper mines, as its promoters hope. Mine backers have said such an expansion would get a closer environmental review later if they pursue it. Scientists have raised concerns that even the smaller mine could have wide impacts, because the resilience of the salmon runs hinges on access to a wide variety of spawning habitats. Environmental groups have vowed to file lawsuits to block the project. 90% —Accuracy of a new artificial intelligence system trained to identify individual weaver birds, which human birders generally cannot tell apart unless they are tagged ( Methods in Ecology and Evolution ). ### Planetary science China's first independent mission to Mars blasted off from the Xichang Satellite Launch Center on 23 July. To arrive in February 2021, Tianwen-1, a “quest for heavenly truth,” comprises an orbiter, lander, and rover. Only the United States and the Soviet Union have successfully landed on Mars. Instruments on the three Tianwen-1 craft will study the planet's magnetic field and atmosphere, map its surface, and characterize its geology. Tianwen-1 is the second in a trio of fresh martian missions: The United Arab Emirates launched its Hope orbiter on 19 July, and NASA planned to launch its Perseverance rover as early as 30 July, after Science went to press. ### Funding A bill in France would increase research spending over the next 10 years and add tenure-track faculty positions, a novelty in France. But critics say the plan's increases would be too small and slow. By 2030, the annual public research budget would rise by about one-third, to €20 billion, toward a goal of lifting overall R&D spending from 2.2% of gross domestic product to 3%. The National Research Agency, which funds researchers through competitive calls, would get €1 billion more over 7 years, reaching about €1.7 billion in 2027, to help raise its grant success rates from 16% to a target of 30%. The new, nonpermanent tenure-track positions would complement the permanent entry-level research positions traditionally offered by the French system, but critics fear the growth may lead to a decline in the permanent ones. Parliament is expected to approve the bill. ### Drug trials A monoclonal antibody given to babies has strongly protected them from severe disease caused by respiratory syncytial virus (RSV), a leading cause of infant death. As reported this week in The New England Journal of Medicine , a placebo-controlled study of nearly 1500 babies born preterm—who are at higher risk of severe symptoms of RSV—in 23 countries found that a single injection of the antibody before RSV season starts in the fall led to 78.4% fewer hospitalizations for lower respiratory infections associated with the disease. The antibody, being developed by AstraZeneca and Sanofi Pasteur, could replace one now on the market that is rarely used. (It is recommended only for infants at highest risk, requires five shots, and is very expensive.) The companies plan to seek regulatory approval of the new prophylaxis if larger studies now underway in preterm and full-term infants confirm that it is safe and effective. ### Graduate studies The American Astronomical Society last week launched the Astronomy Genealogy Project, which maps 5000 astronomers to their academic “descendants”—the 28,000 doctorate recipients they supervised. The discipline's family tree, at astrogen.aas.org, stretches back to 1766, but half of the listed doctorates were awarded since 2002. Organizers hope the data will help historians and sociologists of science analyze patterns across countries, universities, and subfields. U.S. universities awarded slightly more than half of the doctorates listed, and about two-thirds of the theses are online. ### Climate Environmental groups last week denounced as weak a plan announced by the U.S. Environmental Protection Agency (EPA) to limit greenhouse gas emissions from aircraft. The new standard would match an existing one adopted in 2016 by a U.N. body, the International Civil Aviation Organization (ICAO), that required emissions cuts by 2028. But recently manufactured planes already meet the standard, and EPA conceded its new rule would not reduce overall airplane emissions. Manufacturers have supported such a U.S. regulation to help them meet ICAO certification requirements. IACO has predicted that even under its standard, airplane emissions will grow by at least 3% a year globally. U.S. aviation accounts for 3% of the country's greenhouse gas emissions. ### Extreme life Bacteria from seafloor sediments buried 101 million years ago have been grown in the lab, raising the possibility they are as old as their muddy home. They had somehow survived in an area of the Pacific Ocean almost devoid of organic matter or other nutrients most bacteria need, although the sediments recovered do contain oxygen, the researchers report in Nature Communications . The finding pushes back the documented age of bacteria living in marine sediment from 15 million years and provides new insights on the limits of life under extreme conditions. A team led by researchers from the Japan Agency for Marine-Earth Science and Technology harvested the microbes from core samples drilled up to 5700 meters below sea level and took precautions against contaminating them with modern bacteria. The group argues the microbes likely didn't have enough food to keep replicating, and instead may have survived for eons without dividing by repairing age-related cellular damage. The microbes identified are known members of more than eight bacterial groups, many of which are commonly found elsewhere on Earth. ### Biotechnology Scientists announced last week that they used CRISPR gene editing to modify a cow embryo so that the resulting calf, named Cosmo, should produce more offspring bearing male traits. Bulls are 15% more efficient than cows at converting feed into weight gain, so the new method may allow cattle farmers to raise fewer cattle, benefiting the environment, say the researchers at the University of California, Davis. The researchers inserted a gene called SRY , which initiates male development and is normally found on the male sex chromosome, into an embryo's chromosome 17. Next, the researchers plan to determine whether Cosmo's offspring that inherit the SRY gene look and grow like males. Fifty percent of the calf's progeny will naturally be male; another 25% will be genetically female but will carry the SRY gene. ### Conservation Florida's governor this month signed a bill to establish a 162,000-hectare marine sanctuary in the Gulf of Mexico and protect one of the state's last remaining stretches of seagrass. Florida's coast boasts the most continuous expanse of seagrass beds in the United States, but these diverse habitats, home to blue crabs and manatees, have been damaged by nutrient-driven algal blooms and boat propellers. Authorities plan to create a management plan for the new Nature Coast Aquatic Preserve to balance protection with ecotourism, boating, and fishing. ### A magic ride for science In 1984, artist Bruce Degen met writer Joanna Cole at a publisher's office in New York City to discuss creating a children's book about science. They went on to collaborate and publish 13 colorful, zany books in The Magic School Bus series, featuring the ebullient, intrepid teacher Ms. Frizzle (above, right), who takes her students on fantastic adventures into the ocean, across the Solar System, and through the human body, for example. Cole died on 12 July at age 75. But the series continues to teach young readers and their parents about the natural world. > Q: Did you expect to create such a legacy? > A: I was in art school doing very serious art, and I realized that, in my heart of hearts, I wanted to do children's books. In the beginning, it was darn hard work. Some book sketch dummies have five layers of rewrites and reillustrations. The first book was a one-book contract to see if this would work. [The reception] was like the world was waiting for somebody to make this happen. People say, “[As a child,] I [used to] read these books, now I read them to my kids!” I could never have imagined it. > Q: Does scientific accuracy get in the way of storytelling? > A: Frequently. You have to tell kids what is true, but you can't give them all the truth—it's too much. For example, the evolution book goes from now [back] to the beginning of the Earth. I [initially] tried to show every era, year, and life form. It was too complicated. So it ended up as a nice, open spiral with a few representations of each era. > Q: Why use the format of adventures? > A: By following the story, it gave kids a mental filing system—they could retrieve and remember information because it was given to them in a memorable trip. ### Dispatches from the pandemic Read additional Science coverage of the pandemic at [sciencemag.org/tags/coronavirus][1]. #### U.S. vaccine efficacy trials begin The first large-scale efficacy trials of COVID-19 vaccines in the United States began last week. On 27 July, the National Institutes of Health, working with Moderna, announced the start of one that aims to recruit 30,000 people. Later that day, a partnership between Pfizer and BioNTech announced separately it was launching a similarly sized study at sites in the United States and elsewhere. Both the Moderna and the Pfizer/BioNTech vaccines contain messenger RNA that prompts cells to make a protein that studs the surface of the COVID-19 virus. If the vaccines work, this viral protein will safely teach the immune system how to battle the virus if a person later is exposed to it. Operation Warp Speed, the Trump administration's push to accelerate development of a COVID-19 vaccine, has committed nearly $3 billion to these two R&D projects, about half its total investment. Other efficacy trials of various COVID-19 vaccines have begun in Brazil, the United Arab Emirates, and the United Kingdom. Results are expected in late fall at the earliest. #### CDC slammed over school rules Guidance issued by the U.S. Centers for Disease Control and Prevention (CDC) last week for safely reopening schools downplays risks that teachers, other staff members, and students will spread or contract COVID-19, many public health specialists say. Provoking claims that CDC's advice had been politicized, the agency revised an earlier draft that President Donald Trump had panned as “very tough and expensive.” The nonbinding recommendations, released 23 July, emphasize the social and developmental benefits of in-person schooling and highlight that young children are at low risk for contracting the disease and transmitting the virus that causes it. The document also recommends against screening students for symptoms. Nevertheless, the Trump administration did advise communities with high infection rates to consider not beginning in-person classes. Large school districts, such as those in Atlanta, Houston, Los Angeles, and San Diego, have already announced that they will begin the 2020–21 school year with online instruction only. #### Anti-Fauci TV segment canceled Following heavy criticism from scientists and others, Sinclair Broadcast Corp. this week canceled plans for its chain of local TV stations to air a segment featuring widely challenged accusations that Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, intentionally created the virus responsible for COVID-19 and sent it to China. Fauci has helped lead the U.S. effort to control the pandemic despite tangling with President Donald Trump. The allegation came from Judy Mikovits, a virologist and antivaccine activist who appears in a documentary about the coronavirus that was also widely debunked as false and misleading (). The Sinclair segment, a new interview with Mikovits, was available online until the company pulled it for review on 25 July, after Media Matters reported its existence. Sinclair announced on 27 July that it would not air the segment on the nearly 200 TV stations it owns or operates in 89 U.S. markets—but not before one in Charleston, West Virginia, had broadcast it. [1]: http://sciencemag.org/tags/coronavirus

Endangered sperm whales are less likely to forage for food at dawn in some areas of the Mediterranean, underwater robotic equipment has revealed. Unmanned underwater robots equipped with acoustic monitors recorded the sperm whale sounds over several months and thousands of miles of ocean. Sperm whales emit distinct'clicks' to sense objects from reflected sound waves – a process called echo-location – and social interaction purposes. The recordings confirmed the whales' widespread presence in the north-western Mediterranean Sea – especially in the Gulf of Lion, just of the south coast of France. However, in the Gulf of Lion, click recordings showed a clear pattern of decreased foraging efforts, indicated by fewer clicks, at dawn.

Most of the existing coastal flood Forecast and Early-Warning Systems do not model the flood, but instead, rely on the prediction of hydrodynamic conditions at the coast and on expert judgment. Recent scientific contributions are now capable to precisely model flood events, even in situations where wave overtopping plays a significant role. Such models are nevertheless costly-to-evaluate and surrogate ones need to be exploited for substantial computational savings. For the latter models, the hydro-meteorological forcing conditions (inputs) or flood events (outputs) are conveniently parametrised into scalar representations. However, they neglect the fact that inputs are actually functions (more precisely, time series), and that floods spatially propagate inland. Here, we introduce a multi-output Gaussian process model accounting for both criteria. On various examples, we test its versatility for both learning spatial maps and inferring unobserved ones. We demonstrate that efficient implementations are obtained by considering tensor-structured data and/or sparse-variational approximations. Finally, the proposed framework is applied on a coastal application aiming at predicting flood events. We conclude that accurate predictions are obtained in the order of minutes rather than the couples of days required by dedicated hydrodynamic simulators.

In the mid-nineteen-seventies, the West German Army, the Bundeswehr, built a vast underground bunker near the town of Traben-Trarbach. It was five stories deep, had nearly sixty thousand square feet of floor space, and was designed to withstand a nuclear attack. Eighty days' worth of survival provisions were stored inside, including an emergency power supply and more than a million litres of drinking water. You entered the facility through an air lock; the interior temperature was set to seventy degrees. The walls were concrete, thirty-one inches thick, and some were lined with copper.

Artificial intelligence - or AI for short - is technology that enables a computer to think or act in a more'human' way. It does this by taking in information from its surroundings, and deciding its response based on what it learns or senses. It affects the the way we live, work and have fun in our spare time - and sometimes without us even realising. AI is becoming a bigger part of our lives, as the technology behind it becomes more and more advanced. Machines are improving their ability to'learn' from mistakes and change how they approach a task the next time they try it.