The convergence of the availability of a vast amount of big data, the speed and stretch of cloud computing platforms, and the advancement of sophisticated machine learning algorithms have given birth to an array of innovations in Artificial Intelligence (AI). Other applications that benefit from the implementation of AI systems in the public sector include food supply chain, energy, and environmental management. Indeed, the benefits that AI systems bring to society are grand, and so are the challenges and worries. The evolving technologies learning curve implies miscalculations and mistakes, resulting in unanticipated harmful impacts. We are living in times when it is paramount that the possibility of harm in AI systems has to be recognized and addressed quickly. Thus, identifying the potential risks caused by AI systems means a plan of measures to counteract them has to be adopted as soon as possible.

Modularity is a central principle throughout the design process for cyber-physical systems. Modularity reduces complexity and increases reuse of behavior. In this paper we pose and answer the following question: how can we identify independent `modules' within the structure of reactive control architectures? To this end, we propose a graph-structured control architecture we call a decision structure, and show how it generalises some reactive control architectures which are popular in Artificial Intelligence (AI) and robotics, specifically Teleo-Reactive programs (TRs), Decision Trees (DTs), Behavior Trees (BTs) and Generalised Behavior Trees ($k$-BTs). Inspired by the definition of a module in graph theory, we define modules in decision structures and show how each decision structure possesses a canonical decomposition into its modules. We can naturally characterise each of the BTs, $k$-BTs, DTs and TRs by properties of their module decomposition. This allows us to recognise which decision structures are equivalent to each of these architectures in quadratic time. Our proposed concept of modules extends to formal verification, under any verification scheme capable of verifying a decision structure. Namely, we prove that a modification to a module within a decision structure has no greater flow-on effects than a modification to an individual action within that structure. This enables verification on modules to be done locally and hierarchically, where structures can be verified and then repeatedly locally modified, with modules replaced by modules while preserving correctness. To illustrate the findings, we present an example of a solar-powered drone controlled by a decision structure. We use a Linear Temporal Logic-based verification scheme to verify the correctness of this structure, and then show how one can modify modules while preserving its correctness.

This paper describes two generally applicable approaches towards the significant improvement of the performance of state-of-the-art extractive question answering (EQA) systems. Firstly, contrary to a common belief, it demonstrates that using the objective with independence assumption for span probability $P(a_s,a_e) = P(a_s)P(a_e)$ of span starting at position $a_s$ and ending at position $a_e$ may have adverse effects. Therefore we propose a new compound objective that models joint probability $P(a_s,a_e)$ directly, while still keeping the objective with independency assumption as an auxiliary objective. Our second approach shows the beneficial effect of distantly semi-supervised shared-normalization objective known from (Clark and Gardner, 2017). We show that normalizing over a set of documents similar to the golden passage, and marginalizing over all ground-truth answer string positions leads to the improvement of results from smaller statistical models. Our results are supported via experiments with three QA models (BidAF, BERT, ALBERT) over six datasets. The proposed approaches do not use any additional data. Our code, analysis, pretrained models, and individual results will be available online.

The proliferation of big data has brought an urgent demand for privacy-preserving data publishing. Traditional solutions to this demand have limitations on effectively balancing the tradeoff between privacy and utility of the released data. Thus, the database community and machine learning community have recently studied a new problem of relational data synthesis using generative adversarial networks (GAN) and proposed various algorithms. However, these algorithms are not compared under the same framework and thus it is hard for practitioners to understand GAN's benefits and limitations. To bridge the gaps, we conduct so far the most comprehensive experimental study that investigates applying GAN to relational data synthesis. We introduce a unified GAN-based framework and define a space of design solutions for each component in the framework, including neural network architectures and training strategies. We conduct extensive experiments to explore the design space and compare with traditional data synthesis approaches. Through extensive experiments, we find that GAN is very promising for relational data synthesis, and provide guidance for selecting appropriate design solutions. We also point out limitations of GAN and identify future research directions.

Wide area networking infrastructures (WANs), particularly science and research WANs, are the backbone for moving large volumes of scientific data between experimental facilities and data centers. With demands growing at exponential rates, these networks are struggling to cope with large data volumes, real-time responses, and overall network performance. Network operators are increasingly looking for innovative ways to manage the limited underlying network resources. Forecasting network traffic is a critical capability for proactive resource management, congestion mitigation, and dedicated transfer provisioning. To this end, we propose a nonautoregressive graph-based neural network for multistep network traffic forecasting. Specifically, we develop a dynamic variant of diffusion convolutional recurrent neural networks to forecast traffic in research WANs. We evaluate the efficacy of our approach on real traffic from ESnet, the U.S. Department of Energy's dedicated science network. Our results show that compared to classical forecasting methods, our approach explicitly learns the dynamic nature of spatiotemporal traffic patterns, showing significant improvements in forecasting accuracy. Our technique can surpass existing statistical and deep learning approaches by achieving approximately 20% mean absolute percentage error for multiple hours of forecasts despite dynamic network traffic settings.

We move beyond "Is Machine Learning Useful for Macroeconomic Forecasting?" by adding the "how". The current forecasting literature has focused on matching specific variables and horizons with a particularly successful algorithm. In contrast, we study the usefulness of the underlying features driving ML gains over standard macroeconometric methods. We distinguish four so-called features (nonlinearities, regularization, cross-validation and alternative loss function) and study their behavior in both the data-rich and data-poor environments. To do so, we design experiments that allow to identify the "treatment" effects of interest. We conclude that (i) nonlinearity is the true game changer for macroeconomic prediction, (ii) the standard factor model remains the best regularization, (iii) K-fold cross-validation is the best practice and (iv) the $L_2$ is preferred to the $\bar \epsilon$-insensitive in-sample loss. The forecasting gains of nonlinear techniques are associated with high macroeconomic uncertainty, financial stress and housing bubble bursts. This suggests that Machine Learning is useful for macroeconomic forecasting by mostly capturing important nonlinearities that arise in the context of uncertainty and financial frictions.

For more than four decades, UC San Diego Professor of Physics Patrick H. Diamond and his research group have been advancing fundamental concepts in plasma physics, which is an important aspect of furthering advancements in fusion energy. Most recently, Diamond worked with graduate student Robin Heinonen on a model reduction study that used the Comet supercomputer at the San Diego Supercomputer Center at the University of California San Diego to showcase how machine learning produced a novel model for plasma turbulence. Diamond and Heinonen say that advances in machine learning, such as new deep learning techniques, have provided them with new tools to better understand the self-organization process that emerges from what the researchers term as a seemingly chaotic process. "Turbulence and its transport is chaotic in a sense, but this chaos is ordered and constrained," said Heinonen, who co-authored Turbulence Model Reduction by Deep Learning with Diamond in the academic journal entitled Physical Review E. "Moreover, in certain turbulent systems, the chaos conspires to spontaneously form large, long-lived coherent structures and in many cases, we only have a tenuous understanding of why and now. There are definitely aspects of structure formation and self-organization which we do understand, but it's still an active area of research."

Army researchers recently completed a simulation study where crew members and artificial intelligent agents demonstrated trust and cohesion while working together.U.S. Army Combat Capabilities Development Command's Army Research Laboratory researchers and U.S. Army Military Academy at West Point cadets conducted the study as part of an academic capstone project. It also supports the Army Wingman Joint Capabilities Technology Demonstration and the Army's Next Generation Combat Vehicle mission prioritization."The "Subjective, behavioral, performance, communication and physiological data were collected to identify possible team trust and team cohesion metrics."Researchers "The cadets used the Wingman simulation testbed, which allows a human crew to interact with the actual robotic vehicle autonomy on a realistic gunnery task. They collected informed consent, briefed the participants, and collected questionnaire data, along with timing the event."The "The cadets filled the roles of mobility and lethality operator with me as vehicle commander.

The 75th anniversary of Japan formally surrendering to the U.S. aboard the battleship USS Missouri on Sept. 2, 1945, arrives at a moment when the question of how the war is remembered feels more necessary than ever. Veterans' stories, books, movies and TV shows have kept memories of the war alive for the last 75 years, but how will those stories be told when there are fewer people around who lived through those era-defining years? Recently, some people in younger generations have turned to a perhaps surprising source for World War II stories: video games. Games have become more realistic not only in terms of technological advancements, but also in terms of featuring real people and, at least in terms of blockbuster games like Medal of Honor and Call of Duty, getting input from real experts on military history. For example, the upcoming virtual-reality game Medal of Honor: Above and Beyond will feature documentary shorts, and creators interviewed WWII veterans about their wartime experiences to inform the set, which includes missions across Europe and in Tunisia.

Imagine a world without the internet, a Google search engine, magnetic-resonance imaging, or the Human Genome Project—a sampling of American innovations that, like many scientific tools and research efforts, evolved from U.S. R&D investments. Until Congress recently boosted federal R&D funding as a share of the U.S. economy, such investments had been on the decline—sliding from a high just shy of 1.9% of the gross domestic product in 1964 to 0.62% of the nation's gross domestic product (GDP) in 2018—impeding scientific advances, slowing innovation, and clipping the nation's share of global R&D funding at a time when the country faces challenges, economic and human, triggered by coronavirus disease 2019 (COVID-19). The American Association for the Advancement of Science presented recommendations on how to advance scientific discovery, expand innovation, drive economic advancement, and ensure the scientific community supports opportunities for all at the “Fueling American Innovation and Recovery” hearing held by the U.S. House of Representatives Budget Committee on 8 July. Sudip Parikh, AAAS CEO and executive publisher of the Science family of journals, was among the expert witnesses at the hearing that examined ways to reinvigorate U.S. economic competitiveness, renew federal investment in scientific R&D, and address the global pandemic. Since 1995, the global ranking of U.S. R&D investment as a percentage of its GDP slipped from 4th to 10th place, said House Budget Committee Chairman John Yarmuth during the hearing. The United States now lags behind competitors such as South Korea, Taiwan, Japan, and Germany in R&D investments, said Deborah L. Wince-Smith, president and CEO of the Council on Competitiveness, during her testimony at the House Budget Committee hearing. Yarmuth warned that delays in scientific research projects as well as other economic challenges triggered by COVID-19 could diminish the U.S. position as a leader of global R&D funding. Updating a long-standing framework that guides federal investment in scientific research, ensuring effective coordination of federal responses to the crises facing the nation, and committing to provide scientific evidence to propel racial equity in science and national policy-making formed the core of Parikh's recommendations to the panel. Parikh pointed to research that found U.S. R&D as a share of GDP is well below its historic peak and below the current investment levels of nine other countries. U.S. funded R&D projects drive innovation forward. The crises at hand require a federal investment of 1.9% of GDP, a level that represents an annual funding increase of 11% in scientific R&D, said Parikh. Support for the full spectrum of innovation is needed, including “fundamental science, mission-driven technology, useful knowledge programs to meet local, national and international needs with the federal government as a key partner,” said Parikh. New approaches to R&D funding models and networks should be explored, including “project grants, people-centered grants, teams and hubs, and prizes.” “Broadly, federal research is effective in producing discoveries that lead to high-impact, novel inventions, often in technology areas that have not yet received much industry attention,” said Parikh. AAAS tracks administration and congressional appropriations through the R&D Budget and Policy program, an outreach effort that keeps scientists and policy-makers informed through regular analyses, reports and media outreach. AAAS also engages in advocacy efforts to address executive branch and congressional actions that pose negative consequences for U.S. scientific research activities. AAAS has issued, for instance, statements on the U.S. withdrawal from the World Health Organization, which collaborates across borders to support human health, and an administration proposal to limit the participation of international students at U.S. academic institutions despite a long history of important scientific contributions that foreign national students have made to the U.S. scientific enterprise. “From the beginning, the Trump administration has taken a hard-fiscal line on most research and development programs, favoring Department of Defense technology development and acquisition at the expense of basic and applied research, even Defense research activities,” noted Matthew Hourihan, director of the AAAS R&D Budget and Policy Program. In fiscal year 2017, Congress rejected steep spending cuts outlined in the administration's budget proposal and instead adopted significant spending increases, particularly for the Department of Energy's Office of Science for basic science and R&D programs. By fiscal year 2020, congressional R&D increases stood as a rebuttal to the president's consistent budget reduction proposals that sought more than $12 billion in spending to be shed from federal basic and applied research programs, according to a 2019 analysis by the AAAS R&D Budget and Policy Program. Federal R&D investment supports government laboratories, research activities at federal agencies, academic institutions, and private-sector facilities to drive U.S. scientific advances. The 2020 fiscal year spending package Congress approved dedicated the largest funding increases to the life sciences, particularly for the National Institutes of Health's basic research on human health and related topics, and made low-carbon energy and space exploration programs the second-largest funding recipients, as documented by the AAAS R&D Budget and Policy dashboard that tracks congressional appropriations trends across the scientific enterprise. Despite the rise in the levels of federal R&D funding from 2000 to 2017, “the share of total U.S. R&D funded by the federal government declined from 25% to 22%,” according to “The State of U.S. Science and Engineering,” which highlights the 2020 “Science and Engineering Indicators,” a suite of reports that provide findings on thematic scientific topics. The National Science Board is required to deliver the findings to Congress and the president every 2 years. The business sector and U.S. academic institutions of higher education have stepped up financial support for R&D programs and activities, aware that R&D investments often spark novel scientific knowledge that, in turn, opens new research avenues, contributes to the training of young scientists, and helps fuel the U.S. economy. The business sector plays the most prominent role, having outpaced federal R&D funding to become primarily responsible for the rise in R&D support since 2000. Universities and colleges are the second-largest contributors to R&D and play an important role in the progress of the nation's overall R&D activities by funding more than half of both U.S. basic research and the training of incoming scientists and engineers, according to the 2020 “Science and Engineering Indicators.” Yet, the combination of an overall decline in state support for public universities and colleges, and a leveling off of federal R&D funding for academic institutions at an annual $30 billion, risks a weakening of the United States' standing in innovation, reports suggest. More encouraging, federal science and engineering support for historically black colleges and universities delivered HBCUs a 5.4% R&D funding increase for research and experimental development, according to the National Center for Science and Engineering Statistics. Joining Parikh and Wince-Smith at the House hearing were two other expert witnesses, including Paul Romer, a Nobel-Prize winning economist and professor at New York University. Romer called on U.S. universities to give talented science and technology students a larger voice and leeway to pursue their own innovative ideas, not just those of professors. He said science and engineering degree fellowships should be granted to highly talented undergraduates to help expand U.S. innovation. Economist Willy Shih, a professor at Harvard Business School, said the COVID-19 pandemic has exposed U.S. reliance on other nations for equipment, devices, and pharmaceuticals, a reality that requires expanded federal investment in basic scientific research and direct stimulus spending for technology investments. “If the United States does not make needed investments in its future, increase its scope and rate of innovation, its fundamental capacity to grow its economy, create jobs, maintain national security, solve societal challenges and provide a social safety net, it will continue to erode—and its geopolitical leadership will be at increasing risk,” said Wince-Smith.