A key goal of computer vision is to recover the underlying 3D structure from 2D observations of the world. In this paper we learn strong deep generative models of 3D structures, and recover these structures from 3D and 2D images via probabilistic inference. We demonstrate high-quality samples and report log-likelihoods on several datasets, including ShapeNet [2], and establish the first benchmarks in the literature. We also show how these models and their inference networks can be trained end-to-end from 2D images. This demonstrates for the first time the feasibility of learning to infer 3D representations of the world in a purely unsupervised manner.

Earlier this week, I found myself answering a question from a new colleague at Finning International that relates both to the research I do in the iSchool at the University of British Columbia, as well as the analytics, engineering & technology work that I lead at Finning. The questions were simple: 1) What is artificial intelligence? As I sat to reflect last evening, it dawned on me that taking time to craft a clear answer to these questions might be extremely beneficial for many. Analytics, data science, and predictive intelligence are hot topics in many communities and business areas. And yet, despite this interest, few folks I have talked to have a clear understanding of the history of the discipline; one, that frames much of the work currently going on within the space.

So you know the Bayes rule. How does it relate to machine learning? It can be quite difficult to grasp how the puzzle pieces fit together - we know it took us a while. This article is an introduction we wish we had back then. While we have some grasp on the matter, we're not experts, so the following might contain inaccuracies or even outright errors. Feel free to point them out, either in the comments or privately.

The artificial intelligence, dubbed ALPHA, was the victor in that simulated scenario, and according to Lee, is "the most aggressive, responsive, dynamic and credible AI I've seen to date." Details on ALPHA -- a significant breakthrough in the application of what's called genetic-fuzzy systems are published in the most-recent issue of the Journal of Defense Management, as this application is specifically designed for use with Unmanned Combat Aerial Vehicles (UCAVs) in simulated air-combat missions for research purposes. The tools used to create ALPHA as well as the ALPHA project have been developed by Psibernetix, Inc., recently founded by UC College of Engineering and Applied Science 2015 doctoral graduate Nick Ernest, now president and CEO of the firm; as well as David Carroll, programming lead, Psibernetix, Inc.; with supporting technologies and research from Gene Lee; Kelly Cohen, UC aerospace professor; Tim Arnett, UC aerospace doctoral student; and Air Force Research Laboratory sponsors. ALPHA is currently viewed as a research tool for manned and unmanned teaming in a simulation environment. In its earliest iterations, ALPHA consistently outperformed a baseline computer program previously used by the Air Force Research Lab for research.

Aptly name artificial intelligence, ALPHA recently beat a veteran aerial combat expert in a high-fidelity combat simulator. News of an AI beating a highly skilled combat pilot has caused ripples, not only across the artificial intelligence industry, but also the entire tech, social media community. This, a landmark achievement in what's known as genetic-fuzzy systems, was the brainchild of a collaboration between AI development firm - Psibernetix, U.S. Air Force, and a team of scientists from University of Cincinnati. Gene Lee, who is a retired U.S. Air Force Colonel with oodles of experience as an instructor as well as an Air Battle Manager, lost to AI ALPHA, after sparring in what was an action-packed air combat simulation. Lee described ALPHA as, "the most aggressive, responsive, dynamic and credible AI" he has ever seen.

ALPHA, whose destiny was once just to help train fighter pilots in simulated exercises, has proved to be so good that the Air Force Research Laboratory (AFRL) is investigating whether the bot could provide the AI blueprint for combat drones in the future, potentially flying missions alongside human pilots. "ALPHA is an incredibly difficult opponent to face," its developer Nick Earnest, CEO of Psibernetix told Digital Trends. "Even flying against other pilots when ALPHA has severe handicaps to a number of its systems -- including speed, turning, missile capability and sensors -- it is able to win," Earnest said. "There is additional work to be done to both increase ALPHA's capabilities and improve its model fidelity -- but these results represent a significant breakthrough," said Earnest, University of Cincinnati graduate said. ALPHA was created using a Genetic Fuzzy Tree (GFT) system that uses genetic algorithms to train independent and interconnected systems.

We derive a new discrepancy statistic for measuring differences between two probability distributions based on combining Stein's identity with the reproducing kernel Hilbert space theory. We apply our result to test how well a probabilistic model fits a set of observations, and derive a new class of powerful goodness-of-fit tests that are widely applicable for complex and high dimensional distributions, even for those with computationally intractable normalization constants. Both theoretical and empirical properties of our methods are studied thoroughly.

In the future, the US Air Force hopes to have armed drones flying in formation with human pilots, responding to their verbal and digital commands to fight the enemy and strike targets. That would require an artificial intelligence capable of interpreting commands and applying knowledge of combat tactics--something that is already being proven in a project funded by the Air Force Research Lab. ALPHA, an artificial intelligence trained by a retired Air Force expert in air combat, was originally developed as what amounts to ultimate video game AI--an autonomous simulated enemy for use in training fighter pilots. The AI is so good that it has consistently beaten human pilots in simulated air combat--even when heavily handicapped by simulated physics. And now AFRL is investigating using ALPHA as the AI for Unmanned Combat Aerial Vehicles (UCAVs) in the physical world, potentially flying missions alongside human pilots.

Is AI ready for primetime? In a recent interview with Charlie Rose, he stated that machine learning showed great promise for cybersecurity, but that the necessary technology was probably five years out. If machine learning is currently so successful in other areas of society, why isn't it ready for cybersecurity? Machine learning is a subset of Artificial Intelligence, a field of computer science that started in 1958 when Marvin Minsky founded the Artificial Intelligence lab. Everyone, including DARPA, was pouring money into it.

Approximate Bayesian Computation (ABC) is a framework for performing likelihood-free posterior inference for simulation models. Stochastic Variational inference (SVI) is an appealing alternative to the inefficient sampling approaches commonly used in ABC. However, SVI is highly sensitive to the variance of the gradient estimators, and this problem is exacerbated by approximating the likelihood. We draw upon recent advances in variance reduction for SV and likelihood-free inference using deterministic simulations to produce low variance gradient estimators of the variational lower-bound. By then exploiting automatic differentiation libraries we can avoid nearly all model-specific derivations. We demonstrate performance on three problems and compare to existing SVI algorithms. Our results demonstrate the correctness and efficiency of our algorithm.