Developing policy informed by science and technology is now more complex than ever. Policymakers must address supply chains, climate change, inequality, technological breakthroughs, misinformation and more. Using artificial intelligence (AI) to mine the literature could put policymaking on a sounder footing. Advanced big-data and natural-language-processing models enable decision makers to look beyond conventional indicators and expert discussions. Millions of scientific articles, patents and market reports can be readily analysed to identify megatrends or fading topics, and to provide predictive opportunities (see go.nature.com/31snkp5). Machine learning can create maps of national competencies and centres of excellence of science and technology.

We introduce the eigentask framework for lifelong learning. An eigentask is a pairing of a skill that solves a set of related tasks, paired with a generative model that can sample from the skill's input space. The framework extends generative replay approaches, which have mainly been used to avoid catastrophic forgetting, to also address other lifelong learning goals such as forward knowledge transfer. We propose a wake-sleep cycle of alternating task learning and knowledge consolidation for learning in our framework, and instantiate it for lifelong supervised learning and lifelong RL. We achieve improved performance over the state-of-the-art in supervised continual learning, and show evidence of forward knowledge transfer in a lifelong RL application in the game Starcraft2.

Inspired by e-participation systems, in this paper we propose a new model to represent human debates and methods to obtain collective conclusions from them. This model overcomes drawbacks of existing approaches by allowing users to introduce new pieces of information into the discussion, to relate them to existing pieces, and also to express their opinion on the pieces proposed by other users. In addition, our model does not assume that users' opinions are rational in order to extract information from it, an assumption that significantly limits current approaches. Instead, we define a weaker notion of rationality that characterises coherent opinions, and we consider different scenarios based on the coherence of individual opinions and the level of consensus that users have on the debate structure. Considering these two factors, we analyse the outcomes of different opinion aggregation functions that compute a collective decision based on the individual opinions and the debate structure. In particular, we demonstrate that aggregated opinions can be coherent even if there is a lack of consensus and individual opinions are not coherent. We conclude our analysis with a computational evaluation demonstrating that collective opinions can be computed efficiently for real-sized debates.

Clustering is a foundational problem in machine learning with numerous applications. As machine learning increases in ubiquity as a backend for automated systems, concerns about fairness arise. Much of the current literature on fairness deals with discrimination against protected classes in supervised learning (group fairness). We define a different notion of fair clustering wherein the probability that two points (or a community of points) become separated is bounded by an increasing function of their pairwise distance (or community diameter). We capture the situation where data points represent people who gain some benefit from being clustered together. Unfairness arises when certain points are deterministically separated, either arbitrarily or by someone who intends to harm them as in the case of gerrymandering election districts. In response, we formally define two new types of fairness in the clustering setting, pairwise fairness and community preservation. To explore the practicality of our fairness goals, we devise an approach for extending existing $k$-center algorithms to satisfy these fairness constraints. Analysis of this approach proves that reasonable approximations can be achieved while maintaining fairness. In experiments, we compare the effectiveness of our approach to classical $k$-center algorithms/heuristics and explore the tradeoff between optimal clustering and fairness.

Despite the success of existing tensor factorization methods, most of them conduct a multilinear decomposition, and rarely exploit powerful modeling frameworks, like deep neural networks, to capture a variety of complicated interactions in data. More important, for highly expressive, deep factorization, we lack an effective approach to handle streaming data, which are ubiquitous in real-world applications. To address these issues, we propose SPIDER, a Streaming ProbabilistIc Deep tEnsoR factorization method. We first use Bayesian neural networks (NNs) to construct a deep tensor factorization model. We assign a spike-and-slab prior over the NN weights to encourage sparsity and prevent overfitting. We then use Taylor expansions and moment matching to approximate the posterior of the NN output and calculate the running model evidence, based on which we develop an efficient streaming posterior inference algorithm in the assumed-density-filtering and expectation propagation framework. Our algorithm provides responsive incremental updates for the posterior of the latent factors and NN weights upon receiving new tensor entries, and meanwhile select and inhibit redundant/useless weights. We show the advantages of our approach in four real-world applications.

Staying on top of cybersecurity trends and threats is almost a full-time job. The industry expands every day as new applications come online for the internet of things, machine learning and artificial intelligence. But cybercriminals are innovating, too. Not a day passes without news of a major ransomware attack or phishing scheme. Businesses large and small now must learn the intricacies of data breaches and the dark web, negotiating with hackers, buying cyberinsurance and training employees on the best cyberhygiene.

Artificial intelligence is the major buzzword in federal IT these days, the way that cloud once was. It's easy to see why. There is booming investment in AI in the private sector, and various agencies across the government are experimenting with AI to achieve their missions. The National Oceanic and Atmospheric Administration is working with Microsoft to use AI and cloud technology to more easily and accurately identify animals and population counts of endangered species. NASA is ramping up the use of AI throughout its operations, from conducting basic financial operations to finding extra radio frequencies aboard the International Space Station.

The term'covidiot' is a coronavirus-era slang term for someone who ignores recommendations to limit the spread of the deadly disease – and a new study reveals what makes these people dismiss the warnings. Researchers found that whether or not an individual decides to follow social distancing depends on how much information their working memory can store, which determines mental abilities such as intelligence. Following a survey of 850 Americans, the team discovered that those with more working memory capacity were more likely to comply with recommendations during the early stage of the outbreak. The findings suggest that policy makers need promote compliance behaviors, such as wearing a mask, based on individuals' general cognitive abilities to avoid effortful decisions. The coronavirus began spread across the US earlier this year and when it gained more traction, the Centers for Disease Control and Prevention (CDC) released a list of recommendations aimed at limiting the spread of the virus.

The first five months of 2020 sent a parade of "wicked problems" around the globe, including a plague of locusts in Asia and Africa, bushfires in Australia and, of course, the COVID-19 pandemic. Wicked problems can be defined as problems that no one knows how to solve without creating further problems. We struggle to mitigate them because they transcend borders and generations. During and after World War II, policymakers also confronted significant problems, such as how to keep the peace, encourage recovery and prevent starvation. They tackled these problems by creating collaborative institutions and rules, and by providing generous aid and technical assistance.

How do planetary systems--like our solar system or multi-planet systems around other stars--organize themselves? Of all of the possible ways planets could orbit, how many configurations will remain stable over the billions of years of a star's life cycle? Rejecting the large range of unstable possibilities--all the configurations that would lead to collisions--would leave behind a sharper view of planetary systems around other stars, but it's not as easy as it sounds. "Separating the stable from the unstable configurations turns out to be a fascinating and brutally hard problem," said Daniel Tamayo, a NASA Hubble Fellowship Program Sagan Fellow in astrophysical sciences at Princeton. To make sure a planetary system is stable, astronomers need to calculate the motions of multiple interacting planets over billions of years and check each possible configuration for stability--a computationally prohibitive undertaking.