AI holds the potential to replace humans for tactical tasks in military operations beyond current applications such as navigation assistance. For example, in the US, the Defense Advanced Research Projects Agency (DARPA) recently held the final round of its AlphaDogfight Trials where an algorithm controlling a simulated F-16 fighter was pitted against an Air Force pilot in virtual aerial combat. The algorithm won by 5-0. So what does this mean for the future of military operations? The agency's deputy director remarked that these tools are now'ready for weapons systems designers to be in the toolbox'.

The teams had to start from the ground up to teach their AIs how to fly a fighter jet. Lee Ritholtz, director and chief architect of AI, from Lockheed Martin, said: "You don't have to teach a human [that] it shouldn't crash into the ground. "They have basic instincts that the algorithm doesn't have. "That means dying a lot. For Lockheed Martin, it took several servers running trial-and-error dogfights around the clock to come up with its final AI, a piece of software capable of being run on a single graphics card. The winning team's AI had been through more than 4 billion simulations.

Supply Lines is a daily newsletter that tracks Covid-19's impact on trade. Sign up here, and subscribe to our Covid-19 podcast for the latest news and analysis on the pandemic. China imposed new restrictions on the export of artificial intelligence technologies, signaling it may step in to block Beijing-based ByteDance Ltd. from selling the U.S. operations of its TikTok short-video app. AI interface technologies such as speech and text recognition, and those that analyze data to make personalized content recommendations, were added to a revised list of export-control products published on the Ministry of Commerce's website late Friday. Government permits will be required for overseas transfers to "safeguard national economic security," it said.

A new work permit introduced by Japan for overseas workers to help alleviate chronic labor shortages in certain industries has made an unexpectedly poor start, with only 3,987 of them obtaining the "specific skills visa" in the first year of the program, or less than 10 percent of the government's target. The weak start has exposed the insufficient preparations for the program launched in April 2019. With the impact of the COVID-19 pandemic unavoidable, the new system is already at a turning point. There are two ways for foreign workers to obtain the new visa. One is to pass an exam that measures Japanese language proficiency and the skills needed for the industry in which the applicant wants to work. In the other route, for the Type 1 visa for less sophisticated jobs, people who went through technical training in Japan for three years or more can change their visa status to specific skills without taking the exam.

The NHS is set to receive millions of pounds for artificial intelligence upgrades as it works to clear a backlog of millions of cancer patients. It comes after Health Secretary Matt Hancock admitted'some cancer treatment had to stop' and couldn't rule out having to cancel operations again if coronavirus cases surged. The NHS is working to clear a backlog of potentially millions of patients after either operations were cancelled or people were deterred from going to hospital as the UK entered lockdown. Cancer Research UK said as a result of the pandemic there were 2.4 million patients waiting for cancer screening, further treatment or treatment at the end of May. But the Government said a £50 million funding boost to three digital pathology centres, based in London, Coventry and Leeds, would now lead to a'faster and more accurate' diagnosis for millions of cancer patients.

The U.S. Department of Defense (DoD) has invested in the development of technologies that allow the human brain to communicate directly with machines, including the development of implantable neural interfaces able to transfer data between the human brain and the digital world. This technology, known as brain-computer interface (BCI), may eventually be used to monitor a soldier's cognitive workload, control a drone swarm, or link with a prosthetic, among other examples. Further technological advances could support human-machine decisionmaking, human-to-human communication, system control, performance enhancement and monitoring, and training. However, numerous policy, safety, legal, and ethical issues should be evaluated before the technology is widely deployed. With this report, the authors developed a methodology for studying potential applications for emerging technology. This included developing a national security game to explore the use of BCI in combat scenarios; convening experts in military operations, human performance, and neurology to explore how the technology might affect military tactics, which aspects may be most beneficial, and which aspects might present risks; and offering recommendations to policymakers.

Recent successes combine reinforcement learning algorithms and deep neural networks, despite reinforcement learning not being widely applied to robotics and real world scenarios. This can be attributed to the fact that current state-of-the-art, end-to-end reinforcement learning approaches still require thousands or millions of data samples to converge to a satisfactory policy and are subject to catastrophic failures during training. Conversely, in real world scenarios and after just a few data samples, humans are able to either provide demonstrations of the task, intervene to prevent catastrophic actions, or simply evaluate if the policy is performing correctly. This research investigates how to integrate these human interaction modalities to the reinforcement learning loop, increasing sample efficiency and enabling real-time reinforcement learning in robotics and real world scenarios. This novel theoretical foundation is called Cycle-of-Learning, a reference to how different human interaction modalities, namely, task demonstration, intervention, and evaluation, are cycled and combined to reinforcement learning algorithms. Results presented in this work show that the reward signal that is learned based upon human interaction accelerates the rate of learning of reinforcement learning algorithms and that learning from a combination of human demonstrations and interventions is faster and more sample efficient when compared to traditional supervised learning algorithms. Finally, Cycle-of-Learning develops an effective transition between policies learned using human demonstrations and interventions to reinforcement learning. The theoretical foundation developed by this research opens new research paths to human-agent teaming scenarios where autonomous agents are able to learn from human teammates and adapt to mission performance metrics in real-time and in real world scenarios.

Trustworthiness is a central requirement for the acceptance and success of human-centered artificial intelligence (AI). To deem an AI system as trustworthy, it is crucial to assess its behaviour and characteristics against a gold standard of Trustworthy AI, consisting of guidelines, requirements, or only expectations. While AI systems are highly complex, their implementations are still based on software. The software engineering community has a long-established toolbox for the assessment of software systems, especially in the context of software testing. In this paper, we argue for the application of software engineering and testing practices for the assessment of trustworthy AI. We make the connection between the seven key requirements as defined by the European Commission's AI high-level expert group and established procedures from software engineering and raise questions for future work.

The adversarial vulnerability of deep networks has spurred the interest of researchers worldwide. Unsurprisingly, like images, adversarial examples also translate to time-series data as they are an inherent weakness of the model itself rather than the modality. Several attempts have been made to defend against these adversarial attacks, particularly for the visual modality. In this paper, we perform detailed benchmarking of well-proven adversarial defense methodologies on time-series data. We restrict ourselves to the $L_{\infty}$ threat model. We also explore the trade-off between smoothness and clean accuracy for regularization-based defenses to better understand the trade-offs that they offer. Our analysis shows that the explored adversarial defenses offer robustness against both strong white-box as well as black-box attacks. This paves the way for future research in the direction of adversarial attacks and defenses, particularly for time-series data.

Spectral methods have gained a lot of recent attention due to the simplicity of their implementation and their solid mathematical background. We revisit spectral graph clustering, and reformulate in the $p$-norm the continuous problem of minimizing the graph Laplacian Rayleigh quotient. The value of $p \in (1,2]$ is reduced, promoting sparser solution vectors that correspond to optimal clusters as $p$ approaches one. The computation of multiple $p$-eigenvectors of the graph $p$-Laplacian, a nonlinear generalization of the standard graph Laplacian, is achieved by the minimization of our objective function on the Grassmann manifold, hence ensuring the enforcement of the orthogonality constraint between them. Our approach attempts to bridge the fields of graph clustering and nonlinear numerical optimization, and employs a robust algorithm to obtain clusters of high quality. The benefits of the suggested method are demonstrated in a plethora of artificial and real-world graphs. Our results are compared against standard spectral clustering methods and the current state-of-the-art algorithm for clustering using the graph $p$-Laplacian variant.