Influence functions efficiently estimate the effect of removing a single training data point on a model's learned parameters. While influence estimates align well with leave-one-out retraining for linear models, recent works have shown this alignment is often poor in neural networks. In this work, we investigate the specific factors that cause this discrepancy by decomposing it into five separate terms. We study the contributions of each term on a variety of architectures and datasets and how they vary with factors such as network width and training time. While practical influence function estimates may be a poor match to leave-one-out retraining for nonlinear networks, we show they are often a good approximation to a different object we term the proximal Bregman response function (PBRF). Since the PBRF can still be used to answer many of the questions motivating influence functions, such as identifying influential or mislabeled examples, our results suggest that current algorithms for influence function estimation give more informative results than previous error analyses would suggest.

Recent AI algorithms are blackbox models whose decisions are difficult to interpret. eXplainable AI (XAI) seeks to address lack of AI interpretability and trust by explaining to customers their AI decision, e.g., decision to reject a loan application. The common wisdom is that regulating AI by mandating fully transparent XAI leads to greater social welfare. This paper challenges this notion through a game theoretic model for a policy-maker who maximizes social welfare, firms in a duopoly competition that maximize profits, and heterogenous consumers. The results show that XAI regulation may be redundant. In fact, mandating fully transparent XAI may make firms and customers worse off. This reveals a trade-off between maximizing welfare and receiving explainable AI outputs. We also discuss managerial implications for policy-maker and firms.

Sharjah: Artificial Intelligence experts and professionals from government and private sector entities who are integrating AI systems in their operations have shared insights and real-life examples of the transformative power of AI in driving the economic growth of businesses in various sectors during a forum organised by the Sharjah Entrepreneurship Center (Sheraa) recently at its headquarters. The Sheraa AI Forum was held in the presence of HE Major General Saif Al Zari Al Shamsi, Commander-in-Chief of Sharjah Police; H.E. Mohammed Bin Taliah, Chief of Government Services of the UAE Government; Najla Al Midfa, CEO of Sheraa, and a representative of the Minister of State for Artificial Intelligence office, Sharjah Police General Managers, Minister's Office of Artificial Intelligence representative and heads of companies and executives from the government and private sectors. The forum also brought together 150 Emirati youth, entrepreneurs and tech founders to shine light on best practices in the field and the opening of new investment opportunities in the public and private sectors with the ongoing adoption of advanced technologies. During the event, Sharjah Police and Sheraa startups shared their experiences in utilising AI in their products and solutions, noting the transformative power of AI in driving the economic growth of businesses in various sectors. Sheraa AI Forum, which is aligned with the UAE National Strategy for Artificial Intelligence 2031, which includes building work teams to enhance artificial intelligence and formulating joint strategic plans to increase the application of AI mechanisms in various sectors.

WASHINGTON – The Biden administration plans next month to broaden curbs on U.S shipments to China of semiconductors used for artificial intelligence and chipmaking tools, several people familiar with the matter said. The Commerce Department intends to publish new regulations based on restrictions communicated in letters earlier this year to three U.S. companies -- KLA Corp, Lam Research Corp and Applied Materials Inc, the people said, speaking on the condition of anonymity. The plan for new rules has not been previously reported. The letters, which the companies publicly acknowledged, forbade them from exporting chipmaking equipment to Chinese factories that produce advanced semiconductors with sub-14 nanometer processes unless the sellers obtain Commerce Department licenses. The rules would also codify restrictions in Commerce Department letters sent to Nvidia Corp and Advanced Micro Devices last month instructing them to halt shipments of several artificial intelligence computing chips to China unless they obtain licenses.

Exygy seeks an enthusiastic, experienced, and creative Full Stack Engineer who is passionate about making a difference in the world with technology. Join our tight-knit and growing team to build a wide variety of high-impact projects with our civic and health sector clients. This is a full-time remote position. As a senior engineer, you'll spend most of your time on multiple client project teams, building new web applications, while supporting and expanding existing projects. You'll work with our cross-functional core team, and our network of remote contributors to define, design, and deliver high-quality web software.

Ensure your team supports all phases of analytic work product development, from the identification of key business questions through data collection and ETL, and from performing analyses and using a wide range of statistical, machine learning, and applied mathematical techniques to delivery insights to decision-makers.

Iran's recent seizure of unmanned US Navy boats shined a light on a pioneering Pentagon program to develop networks of air, surface and underwater drones for patrolling large regions, meshing their surveillance with artificial intelligence. The year-old program operates numerous unmanned surface vessels, or USVs, in the waters around the Arabian peninsula, gathering data and images to be beamed back to collection centers in the Gulf. The program operated without incident until Iranian forces tried to grab three seven-meter Saildrone Explorer USVs in two incidents, on August 29-30 and September 1. In the first, a ship of Iran's Islamic Revolutionary Guard Corps hooked a line to a Saildrone in the Gulf and began towing it away, only releasing it when a US Navy Patrol boat and helicopter sped to the scene. In the second, an Iranian destroyer picked up two Saildrones in the Red Sea, hoisting them aboard.

Abishek Gupta is the founder and principal researcher at the Montreal AI Ethics Institute and senior Responsible AI leader and expert at Boston Consulting Group; Steven Mills is the Global GAMMA Chief AI Ethics Officer at Boston Consulting Group. The Responsible AI (RAI) domain is at an inflection point: We are moving decidedly from principles to practice. As organizations mature their understanding, they are feeling the pressure to act from customer demands and impending regulatory requirements. RAI means developing and operating artificial intelligence systems that align with organizational values and widely accepted standards of right and wrong while achieving transformative business impact. But successfully operationalizing RAI requires a leader with the right mix of knowledge, skills, abilities and experience, and RAI remains a nascent field.

Transformer architectures have led to remarkable progress in many state-of-art applications. However, despite their successes, modern transformers rely on the self-attention mechanism, whose time- and space-complexity is quadratic in the length of the input. Several approaches have been proposed to speed up self-attention mechanisms to achieve sub-quadratic running time; however, the large majority of these works are not accompanied by rigorous error guarantees. In this work, we establish lower bounds on the computational complexity of self-attention in a number of scenarios. We prove that the time complexity of self-attention is necessarily quadratic in the input length, unless the Strong Exponential Time Hypothesis (SETH) is false. This argument holds even if the attention computation is performed only approximately, and for a variety of attention mechanisms. As a complement to our lower bounds, we show that it is indeed possible to approximate dot-product self-attention using finite Taylor series in linear-time, at the cost of having an exponential dependence on the polynomial order.

Despite the prevalence of recent success in learning from static graphs, learning from time-evolving graphs remains an open challenge. In this work, we design new, more stringent evaluation procedures for link prediction specific to dynamic graphs, which reflect real-world considerations, to better compare the strengths and weaknesses of methods. First, we create two visualization techniques to understand the reoccurring patterns of edges over time and show that many edges reoccur at later time steps. Based on this observation, we propose a pure memorization baseline called EdgeBank. EdgeBank achieves surprisingly strong performance across multiple settings because easy negative edges are often used in the current evaluation setting. To evaluate against more difficult negative edges, we introduce two more challenging negative sampling strategies that improve robustness and better match real-world applications. Lastly, we introduce six new dynamic graph datasets from a diverse set of domains missing from current benchmarks, providing new challenges and opportunities for future research. Our code repository is accessible at https://github.com/fpour/DGB.git.