One afternoon not long ago, I sat entering notes into a patient's medical record. She was in her forties, and her labs showed anemia. The causes of anemia range from menstruation to cancer, and so pinpointing the correct underlying diagnosis is critical. Physicians are trained to formulate a full roster of possibilities, known as the differential diagnosis, and then to work down the list systematically. We're taught to cast a wide net--celiac disease, parasitic infections, thalassemia, lead poisoning, liver disease, B12 deficiency, myeloma, sickle-cell disease, G6PD deficiency--because you'll never make a diagnosis if you haven't included it in your differential.

The Home Office plans to use artificial intelligence to reduce the asylum backlog, and is launching a three-day hackathon in the search for quicker ways to process the 138,052 undecided asylum cases. The government is convening academics, tech experts, civil servants and business people to form 15 multidisciplinary teams tasked with brainstorming solutions to the backlog. Teams will be invited to compete to find the most innovative solutions, and will present their ideas to a panel of judges. The winners are expected to meet the prime minister, Rishi Sunak, in Downing Street for a prize-giving ceremony. Inspired by Silicon Valley's approach to problem-solving, the hackathon will take place in London and Peterborough in May.

A massive fire was ignited in the Crimean port city of Sevastopol following a suspected drone attack on a fuel storage tank. The blaze was assigned the highest ranking in terms of how complicated it will be to extinguish, Mikhail Razvozhayev, the Russian-installed governor, wrote on Telegram on Saturday. The fire was still burning but it had been contained and no one was injured. The oil reservoir fire did not cause any casualties and would not hinder fuel supplies in Sevastopol, he said. "The four fuel tanks that were hit, they are practically burnt out already," said Razvozhayev, adding an area of 1,000 square metres (11,000 square feet) had been engulfed in flames.

LONDON/STOCKHOLM – As recently as February, generative AI did not feature prominently in EU lawmakers' plans for regulating generative artificial intelligence technologies such as ChatGPT. The bloc's 108-page proposal for the AI Act, published two years earlier, included only one mention of the word "chatbot." References to AI-generated content largely referred to deepfakes: images or audio designed to impersonate human beings. By mid-April, however, members of European Parliament (MEPs) were racing to update those rules to catch up with an explosion of interest in generative AI, which has provoked awe and anxiety since OpenAI unveiled ChatGPT six months ago. This could be due to a conflict with your ad-blocking or security software.

Gradient-based training in federated learning is known to be vulnerable to faulty/malicious clients, which are often modeled as Byzantine clients. To this end, previous work either makes use of auxiliary data at parameter server to verify the received gradients (e.g., by computing validation error rate) or leverages statistic-based methods (e.g. median and Krum) to identify and remove malicious gradients from Byzantine clients. In this paper, we remark that auxiliary data may not always be available in practice and focus on the statistic-based approach. However, recent work on model poisoning attacks has shown that well-crafted attacks can circumvent most of median- and distance-based statistical defense methods, making malicious gradients indistinguishable from honest ones. To tackle this challenge, we show that the element-wise sign of gradient vector can provide valuable insight in detecting model poisoning attacks. Based on our theoretical analysis of the \textit{Little is Enough} attack, we propose a novel approach called \textit{SignGuard} to enable Byzantine-robust federated learning through collaborative malicious gradient filtering. More precisely, the received gradients are first processed to generate relevant magnitude, sign, and similarity statistics, which are then collaboratively utilized by multiple filters to eliminate malicious gradients before final aggregation. Finally, extensive experiments of image and text classification tasks are conducted under recently proposed attacks and defense strategies. The numerical results demonstrate the effectiveness and superiority of our proposed approach. The code is available at \textit{\url{https://github.com/JianXu95/SignGuard}}

The Sun constantly releases radiation and plasma into the heliosphere. Sporadically, the Sun launches solar eruptions such as flares and coronal mass ejections (CMEs). CMEs carry away a huge amount of mass and magnetic flux with them. An Earth-directed CME can cause serious consequences to the human system. It can destroy power grids/pipelines, satellites, and communications. Therefore, accurately monitoring and predicting CMEs is important to minimize damages to the human system. In this study we propose an ensemble learning approach, named CMETNet, for predicting the arrival time of CMEs from the Sun to the Earth. We collect and integrate eruptive events from two solar cycles, #23 and #24, from 1996 to 2021 with a total of 363 geoeffective CMEs. The data used for making predictions include CME features, solar wind parameters and CME images obtained from the SOHO/LASCO C2 coronagraph. Our ensemble learning framework comprises regression algorithms for numerical data analysis and a convolutional neural network for image processing. Experimental results show that CMETNet performs better than existing machine learning methods reported in the literature, with a Pearson product-moment correlation coefficient of 0.83 and a mean absolute error of 9.75 hours.

Recent advances in adversarial machine learning have shown that defenses considered to be robust are actually susceptible to adversarial attacks which are specifically customized to target their weaknesses. These defenses include Barrage of Random Transforms (BaRT), Friendly Adversarial Training (FAT), Trash is Treasure (TiT) and ensemble models made up of Vision Transformers (ViTs), Big Transfer models and Spiking Neural Networks (SNNs). We first conduct a transferability analysis, to demonstrate the adversarial examples generated by customized attacks on one defense, are not often misclassified by another defense. This finding leads to two important questions. First, how can the low transferability between defenses be utilized in a game theoretic framework to improve the robustness? Second, how can an adversary within this framework develop effective multi-model attacks? In this paper, we provide a game-theoretic framework for ensemble adversarial attacks and defenses. Our framework is called Game theoretic Mixed Experts (GaME). It is designed to find the Mixed-Nash strategy for both a detector based and standard defender, when facing an attacker employing compositional adversarial attacks. We further propose three new attack algorithms, specifically designed to target defenses with randomized transformations, multi-model voting schemes, and adversarial detector architectures. These attacks serve to both strengthen defenses generated by the GaME framework and verify their robustness against unforeseen attacks. Overall, our framework and analyses advance the field of adversarial machine learning by yielding new insights into compositional attack and defense formulations.

It has become common fare in United States political campaigns: advertisements that make sweeping claims of dystopia if the opposing candidate wins. Manipulated, underexposed images and cherry-picked headlines combine to build a crescendo of doom. But in the wake of Tuesday's announcement that Democratic President Joe Biden will run for reelection, an official Republican Party video has stood out for one specific reason: It was generated completely using artificial intelligence (AI) images. The Republican National Committee's embrace of the "transformative technology of our time" is not surprising given the rapid advancement and availability of AI products, said Darrell West, a senior fellow at the Center for Technology Innovation at the Brookings Institution. The Republican Party's use of AI is an early sign of what is likely to come, he told Al Jazeera.

Bipartisan US lawmakers from both chambers of Congress introduced legislation this week that would formally prohibit AI from launching nuclear weapons. Although Department of Defense policy already states that a human must be "in the loop" for such critical decisions, the new bill -- the Block Nuclear Launch by Autonomous Artificial Intelligence Act -- would codify that policy, preventing the use of federal funds for an automated nuclear launch without "meaningful human control." Aiming to protect "future generations from potentially devastating consequences," the bill was introduced by Senator Ed Markey (D-MA) and Representatives Ted Lieu (D-MA), Don Beyer (D-VA) and Ken Buck (R-CO). Senate co-sponsors include Jeff Merkley (D-OR), Bernie Sanders (I-VT), and Elizabeth Warren (D-MA). "As we live in an increasingly digital age, we need to ensure that humans hold the power alone to command, control, and launch nuclear weapons – not robots," said Markey.

Access to the ChatGPT chatbot has been restored in Italy after its maker OpenAI "addressed or clarified" issues raised by Italy's data protection authority, Italian authorities and OpenAI have said. Microsoft Corp-backed OpenAI took ChatGPT offline in Italy last month after the country's Data Protection Authority, also known as Garante, temporarily banned the chatbot and launched a probe into the artificial intelligence application's suspected breach of privacy rules. The Italian Data Protection Authority described its action as provisional "until ChatGPT respects privacy". The watchdog said ChatGPT developer OpenAI had no legal basis to justify "the mass collection and storage of personal data for the purpose of'training' the algorithms underlying the operation of the platform". It further referenced a data breach on March 20 when user conversations and payment information were compromised, a problem the United States firm blamed on a bug.