Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms -- DORM, DORM+, and AdaHedgeD -- arise from a novel reduction of delayed online learning to optimistic online learning that reveals how optimistic hints can mitigate the regret penalty caused by delay. We pair this delay-as-optimism perspective with a new analysis of optimistic learning that exposes its robustness to hinting errors and a new meta-algorithm for learning effective hinting strategies in the presence of delay. We conclude by benchmarking our algorithms on four subseasonal climate forecasting tasks, demonstrating low regret relative to state-of-the-art forecasting models.

The Association for the Advancement of Artificial Intelligence's 2021 Spring Symposium Series was held virtually from March 22-24, 2021. There were ten symposia in the program: Applied AI in Healthcare: Safety, Community, and the Environment, Artificial Intelligence for K-12 Education, Artificial Intelligence for Synthetic Biology, Challenges and Opportunities for Multi-Agent Reinforcement Learning, Combining Machine Learning and Knowledge Engineering, Combining Machine Learning with Physical Sciences, Implementing AI Ethics, Leveraging Systems Engineering to Realize Synergistic AI/Machine-Learning Capabilities, Machine Learning for Mobile Robot Navigation in the Wild, and Survival Prediction: Algorithms, Challenges and Applications. This report contains summaries of all the symposia. The two-day international virtual symposium included invited speakers, presenters of research papers, and breakout discussions from attendees around the world. Registrants were from different countries/cities including the US, Canada, Melbourne, Paris, Berlin, Lisbon, Beijing, Central America, Amsterdam, and Switzerland. We had active discussions about solving health-related, real-world issues in various emerging, ongoing, and underrepresented areas using innovative technologies including Artificial Intelligence and Robotics. We primarily focused on AI-assisted and robot-assisted healthcare, with specific focus on areas of improving safety, the community, and the environment through the latest technological advances in our respective fields. The day was kicked off by Raj Puri, Physician and Director of Strategic Health Initiatives & Innovation at Stanford University spoke about a novel, automated sentinel surveillance system his team built mitigating COVID and its integration into their public-facing dashboard of clinical data and metrics. Selected paper presentations during both days were wide ranging including talks from Oliver Bendel, a Professor from Switzerland and his Swiss colleague, Alina Gasser discussing co-robots in care and support, providing the latest information on technologies relating to human-robot interaction and communication. Yizheng Zhao, Associate Professor at Nanjing University and her colleagues from China discussed views of ontologies with applications to logical difference computation in the healthcare sector. Pooria Ghadiri from McGill University, Montreal, Canada discussed his research relating to AI enhancements in health-care delivery for adolescents with mental health problems in the primary care setting.

Artificial intelligence is also on the advance in IT security. According to a survey of 300 managers, 96 percent reported preparations in their companies for AI-supported IT attacks. In doing so, they partly rely on the help of "defensive AI". The survey was carried out with the assistance of the AI cybersecurity provider Darktrace. A survey of around 200 IT managers in medium-sized companies came to a more differentiated result.

Large-scale pre-trained models (PTMs) such as BERT and GPT have recently achieved great success and become a milestone in the field of artificial intelligence (AI). Owing to sophisticated pre-training objectives and huge model parameters, large-scale PTMs can effectively capture knowledge from massive labeled and unlabeled data. By storing knowledge into huge parameters and fine-tuning on specific tasks, the rich knowledge implicitly encoded in huge parameters can benefit a variety of downstream tasks, which has been extensively demonstrated via experimental verification and empirical analysis. It is now the consensus of the AI community to adopt PTMs as backbone for downstream tasks rather than learning models from scratch. In this paper, we take a deep look into the history of pre-training, especially its special relation with transfer learning and self-supervised learning, to reveal the crucial position of PTMs in the AI development spectrum. Further, we comprehensively review the latest breakthroughs of PTMs. These breakthroughs are driven by the surge of computational power and the increasing availability of data, towards four important directions: designing effective architectures, utilizing rich contexts, improving computational efficiency, and conducting interpretation and theoretical analysis. Finally, we discuss a series of open problems and research directions of PTMs, and hope our view can inspire and advance the future study of PTMs.

Transformers have achieved great success in many artificial intelligence fields, such as natural language processing, computer vision, and audio processing. Therefore, it is natural to attract lots of interest from academic and industry researchers. Up to the present, a great variety of Transformer variants (a.k.a. X-formers) have been proposed, however, a systematic and comprehensive literature review on these Transformer variants is still missing. In this survey, we provide a comprehensive review of various X-formers. We first briefly introduce the vanilla Transformer and then propose a new taxonomy of X-formers. Next, we introduce the various X-formers from three perspectives: architectural modification, pre-training, and applications. Finally, we outline some potential directions for future research.

Generative Adversarial Network, as a promising research direction in the AI community, recently attracts considerable attention due to its ability to generating high-quality realistic data. GANs are a competing game between two neural networks trained in an adversarial manner to reach a Nash equilibrium. Despite the improvement accomplished in GANs in the last years, there remain several issues to solve. In this way, how to tackle these issues and make advances leads to rising research interests. This paper reviews literature that leverages the game theory in GANs and addresses how game models can relieve specific generative models' challenges and improve the GAN's performance. In particular, we firstly review some preliminaries, including the basic GAN model and some game theory backgrounds. After that, we present our taxonomy to summarize the state-of-the-art solutions into three significant categories: modified game model, modified architecture, and modified learning method. The classification is based on the modifications made in the basic model by the proposed approaches from the game-theoretic perspective. We further classify each category into several subcategories. Following the proposed taxonomy, we explore the main objective of each class and review the recent work in each group. Finally, we discuss the remaining challenges in this field and present the potential future research topics.

Controlling neural network-based models for natural language generation (NLG) has broad applications in numerous areas such as machine translation, document summarization, and dialog systems. Approaches that enable such control in a zero-shot manner would be of great importance as, among other reasons, they remove the need for additional annotated data and training. In this work, we propose novel approaches for controlling encoder-decoder transformer-based NLG models in zero-shot. This is done by introducing three control knobs, namely, attention biasing, decoder mixing, and context augmentation, that are applied to these models at generation time. These knobs control the generation process by directly manipulating trained NLG models (e.g., biasing cross-attention layers) to realize the desired attributes in the generated outputs. We show that not only are these NLG models robust to such manipulations, but also their behavior could be controlled without an impact on their generation performance. These results, to the best of our knowledge, are the first of their kind. Through these control knobs, we also investigate the role of transformer decoder's self-attention module and show strong evidence that its primary role is maintaining fluency of sentences generated by these models. Based on this hypothesis, we show that alternative architectures for transformer decoders could be viable options. We also study how this hypothesis could lead to more efficient ways for training encoder-decoder transformer models.

Leading global technology distributor and solutions provider Avnet Asia will host the "Avnet AI Cloud Exhibition", showcasing innovative technology, applications and solutions in Artificial Intelligence (AI) and machine learning together with its suppliers and partners. With the ability to quickly design, develop and deploy solutions, Avnet can meet the needs of a variety of application scenarios to accelerate the industrialization of artificial intelligence. During this period, Avnet will also hold the "Avnet 2021 Artificial Intelligence Cloud Conference" on June 29, 2021. Joined by developers, engineers, and decision makers in the AI field, the summit will feature cutting-edge technology trends in artificial intelligence and machine learning, and in-depth discussions on the development, future prospects and blueprints for AI to encourage and accelerate innovation. KS Lim, senior director of supplier management at Avnet Asia said, "MarketsandMarkets forecasts the global artificial intelligence (AI) market size to grow to over USD 300 billion by 2026, and the market in Asia Pacific is anticipated to grow at the highest CAGR during the forecast period. As the world's leading technology distributor and solution provider, Avnet has a comprehensive ecosystem that provides customers with end-to-end artificial intelligence and machine learning solutions, reducing the cost and complexity of product development to enable application scenarios. We will continue to work hand in hand with our suppliers and partners to further contribute to the development and maturity of the entire AI ecosystem."

We study the problem of boosting the accuracy of a weak learner in the (distribution-independent) PAC model with Massart noise. In the Massart noise model, the label of each example $x$ is independently misclassified with probability $\eta(x) \leq \eta$, where $\eta<1/2$. The Massart model lies between the random classification noise model and the agnostic model. Our main positive result is the first computationally efficient boosting algorithm in the presence of Massart noise that achieves misclassification error arbitrarily close to $\eta$. Prior to our work, no non-trivial booster was known in this setting. Moreover, we show that this error upper bound is best possible for polynomial-time black-box boosters, under standard cryptographic assumptions. Our upper and lower bounds characterize the complexity of boosting in the distribution-independent PAC model with Massart noise. As a simple application of our positive result, we give the first efficient Massart learner for unions of high-dimensional rectangles.

Nevertheless, the recent advances in machine learning triggered genuine interest, as machine learning offer promising preliminary results and open the way to a wide range of new functions for avionics systems, for instance in the area of autonomous flying. In this paper we investigate on how existing certification and regulation techniques, can (or cannot) handle software development that includes parts obtained by machine learning. Nowadays a large aircraft cockpit offers many avionic complex functions: flight controls, navigation, surveillance, communications, displays... Their design has required a top down iterative approach from aircraft level downward, thus the functions are performed by systems of systems, with each system decomposed into subsystems that may contain a collection of software and hardware items. Therefore, any avionic development considers 3 levels of engineering: (i) Function, (ii) System/Subsystem and (iii) Item. The development process of each engineering level relies on several decades of experience and good practices that keep on being adapted today.