A new report emphasizes why it is urgent that the Department of Defense and Congress work together to modernize the way defense programs and budgets develop, integrate and deploy the latest technologies in support of American national security. Released by the National Security Commission on Artificial Intelligence, a federal body created to review and recommend ways to use artificial intelligence for national security purposes, the report recommends the use of AI to update America's defense plans, predict future threats, deter adversaries and win wars. Because AI will be "incorporated into virtually all future technology," it is easy to recognize that national security threats and opportunities posed by AI should be a catalyst for necessary changes to defense requirements and resourcing processes. In an AI-enabled world, the Defense Department will be unable to modernize the way it recruits talent, trains the force, develops and integrates technology, and funds all of these elements without internal culture shifts and help from Congress. "Unless the requirements, budgeting and acquisition processes are aligned to permit faster and more targeted execution, the U.S. will fail to stay ahead of potential adversaries."

Bayesian inference provides a uniquely rigorous approach to obtain principled justification for uncertainty in predictions, yet it is difficult to articulate suitably general prior belief in the machine learning context, where computational architectures are pure abstractions subject to frequent modifications by practitioners attempting to improve results. Parsimonious inference is an information-theoretic formulation of inference over arbitrary architectures that formalizes Occam's Razor; we prefer simple and sufficient explanations. Our universal hyperprior assigns plausibility to prior descriptions, encoded as sequences of symbols, by expanding on the core relationships between program length, Kolmogorov complexity, and Solomonoff's algorithmic probability. We then cast learning as information minimization over our composite change in belief when an architecture is specified, training data are observed, and model parameters are inferred. By distinguishing model complexity from prediction information, our framework also quantifies the phenomenon of memorization. Although our theory is general, it is most critical when datasets are limited, e.g. small or skewed. We develop novel algorithms for polynomial regression and random forests that are suitable for such data, as demonstrated by our experiments. Our approaches combine efficient encodings with prudent sampling strategies to construct predictive ensembles without cross-validation, thus addressing a fundamental challenge in how to efficiently obtain predictions from data.

Random Forest's performance can be matched by a single slow-growing tree (SGT), which uses a learning rate to tame CART's greedy algorithm. SGT exploits the view that CART is an extreme case of an iterative weighted least square procedure. Moreover, a unifying view of Boosted Trees (BT) and Random Forests (RF) is presented. Greedy ML algorithms' outcomes can be improved using either "slow learning" or diversification. SGT applies the former to estimate a single deep tree, and Booging (bagging stochastic BT with a high learning rate) uses the latter with additive shallow trees. The performance of this tree ensemble quaternity (Booging, BT, SGT, RF) is assessed on simulated and real regression tasks.

We introduce a novel class of stochastic blockmodel for multilayer weighted networks that accounts for the presence of a global ambient noise that governs between-block interactions. We induce a hierarchy of classifications in weighted multilayer networks by assuming that all but one cluster (block) are governed by unique local signals, while a single block is classified as ambient noise, which behaves identically as interactions across differing blocks. Hierarchical variational inference is employed to jointly detect and typologize block-structures as local signals or global noise. These principles are incorporated into novel community detection algorithm called Stochastic Block (with) Ambient Noise Model (SBANM) for multilayer weighted networks. We apply this method to several different domains. We focus on the Philadelphia Neurodevelopmental Cohort to discover communities of subjects that form diagnostic categories relating psychopathological symptoms to psychosis.

The rise of digital payments has caused consequential changes in the financial crime landscape. As a result, traditional fraud detection approaches such as rule-based systems have largely become ineffective. AI and machine learning solutions using graph computing principles have gained significant interest in recent years. Graph-based techniques provide unique solution opportunities for financial crime detection. However, implementing such solutions at industrial-scale in real-time financial transaction processing systems has brought numerous application challenges to light. In this paper, we discuss the implementation difficulties current and next-generation graph solutions face. Furthermore, financial crime and digital payments trends indicate emerging challenges in the continued effectiveness of the detection techniques. We analyze the threat landscape and argue that it provides key insights for developing graph-based solutions.

We present a novel off-policy loss function for learning a transition model in model-based reinforcement learning. Notably, our loss is derived from the off-policy policy evaluation objective with an emphasis on correcting distribution shift. Compared to previous model-based techniques, our approach allows for greater robustness under model misspecification or distribution shift induced by learning/evaluating policies that are distinct from the data-generating policy. We provide a theoretical analysis and show empirical improvements over existing model-based off-policy evaluation methods. We provide further analysis showing our loss can be used for off-policy optimization (OPO) and demonstrate its integration with more recent improvements in OPO.

Proximal Policy Optimization (PPO) is a popular on-policy reinforcement learning algorithm but is significantly less utilized than off-policy learning algorithms in multi-agent problems. In this work, we investigate Multi-Agent PPO (MAPPO), a multi-agent PPO variant which adopts a centralized value function. Using a 1-GPU desktop, we show that MAPPO achieves performance comparable to the state-of-the-art in three popular multi-agent testbeds: the Particle World environments, Starcraft II Micromanagement Tasks, and the Hanabi Challenge, with minimal hyperparameter tuning and without any domain-specific algorithmic modifications or architectures. In the majority of environments, we find that compared to off-policy baselines, MAPPO achieves better or comparable sample complexity as well as substantially faster running time. Finally, we present 5 factors most influential to MAPPO's practical performance with ablation studies.

The progress in Query-focused Multi-Document Summarization (QMDS) has been limited by the lack of sufficient largescale high-quality training datasets. We present two QMDS training datasets, which we construct using two data augmentation methods: (1) transferring the commonly used single-document CNN/Daily Mail summarization dataset to create the QMDSCNN dataset, and (2) mining search-query logs to create the QMDSIR dataset. These two datasets have complementary properties, i.e., QMDSCNN has real summaries but queries are simulated, while QMDSIR has real queries but simulated summaries. To cover both these real summary and query aspects, we build abstractive end-to-end neural network models on the combined datasets that yield new state-of-the-art transfer results on DUC datasets. We also introduce new hierarchical encoders that enable a more efficient encoding of the query together with multiple documents. Empirical results demonstrate that our data augmentation and encoding methods outperform baseline models on automatic metrics, as well as on human evaluations along multiple attributes.

An artificial intelligence commission led by former Google CEO Eric Schmidt is urging the U.S. to boost its AI skills to counter China, including by pursuing "AI-enabled" weapons -- something that Google itself has shied away from on ethical grounds. Schmidt and current executives from Google, Microsoft, Oracle and Amazon are among the 15 members of the National Security Commission on Artificial Intelligence, which released its final report to Congress on Monday. "To win in AI we need more money, more talent, stronger leadership," Schmidt said Monday. The report says that machines that can "perceive, decide, and act more quickly" than humans and with more accuracy are going to be deployed for military purposes -- with or without the involvement of the U.S. and other democracies. It warns against unchecked use of autonomous weapons but expresses opposition to a global ban. It also calls for "wise restraints" on the use of AI tools such as facial recognition that can be used for mass surveillance.

The US military is equipped with a number of stealthy underwater robots to spy on enemies, but these high-tech innovations come years after Charlie the robotic catfish. Developed in the 1990s by the Central Intelligence Agency (CIA), this unmanned underwater vehicle is operated remotely using a line-of-sight audio and fitted with sensors to spy on adversaries, along with collecting water samples. The'catfish' is also designed with a pressure hull, ballast system and communication system in the main part of its body and propulsion system in the tail. Details of Charlie's missions are still classified, but the technology led engineers to design robotic submarines and other aquatic inspired machines to investigate the seas. The robotic fish measures about two feet long and some of its specifications, according to the CIA website, include speed, endurance, depth control and navigational accuracy.