It's a notion marketers will drool over: Imagine if their overloaded job responsibilities could be wiped clean by the use of robots to do tedious marketing tasks for them. Just as we already use programmatic buying and other data-driven marketing tools to simplify difficult and time-consuming processes, artificial intelligence (AI) is viewed by both robotics experts and marketing professionals as a tool to expedite menial content creation in the future. Efforts to bring AI to the mainstream are underway. IBM's Watson AI is already making appearances in ads holding conversations with celebrities. Robotic writing solutions have been used for simple writing tasks, such as recapping sporting events.

I'm often asked about my thoughts on the future. What will transportation be like? What new forms of entertainment will we enjoy? While I have covered these topics in previous articles, it's important to understand that they're all forms of work. So today I want to discuss the future of work -- how technological advancements, namely robotic assistants and tools, as well as tech-enhanced globalization, will affect our daily work flow and the labor market in general.

Human patrols serve as the most direct form of protection of endangered animals, especially in large national parks. However, protection agencies have limited resources for patrols. With support from the National Science Foundation (NSF) and the Army Research Office, researchers are using artificial intelligence (AI) and game theory to solve poaching, illegal logging and other problems worldwide, in collaboration with researchers and conservationists in the U.S., Singapore, Netherlands and Malaysia. "In most parks, ranger patrols are poorly planned, reactive rather than pro-active, and habitual," according to Fei Fang, a Ph.D. candidate in the computer science department at the University of Southern California (USC). Fang is part of an NSF-funded team at USC led by Milind Tambe, professor of computer science and industrial and systems engineering and director of the Teamcore Research Group on Agents and Multiagent Systems.

How we explain reality to ourselves is a construction with many parts. We gather knowledge and generate meaning through our experiences and traditions; from what we learn in school, at work and at home; from how we witness others explaining reality for themselves (on TV, via social media, etc). This narrative that we tell ourselves everyday throughout our entire lives largely defines who we are and how we approach the world. The first time I ran (in Mexico) an adaptation of Stanford's workshop "Makers in Residence" (an intensive 80 hour program for high schoolers on digital fabrication and design thinking which was designed by the Transformative Learning Technology Lab) I was shocked by the comments of participants regarding their place in relation to technology. Most participants were impressed that they were "smarter" than the computers they programmed; when I asked them more about it I started understanding the new narrative that a generation of kids growing up surrounded by digital technology are developing in their heads.

Nintendo's best releases are typically highly polished, moderately zany and, if not downright accessible to a mass audience, at least approachable. They are not just games but showcase titles, works specifically designed to demonstrate the flexibility of Nintendo's own hardware. Its latest, Wii U's "Star Fox Zero," almost fits the profile. The game, equally influenced by high-energy cartoon serials and "Star Wars," is unique, to say the least. "Star Fox" aggressively makes use of the television screen and the Wii U's tablet-like controller, the GamePad, requiring players to navigate between two very distinct points of view.

Tens of thousands of geeked out video game fans from across the globe are flooding this weekend's PAX East festival -- dubbed the "Woodstock for gamers" -- to try their hand at the latest virtual reality tech. "I would never miss PAX for the world," said Carly Monson of Belmont, who turned up to the Boston Convention & Exhibition Center decked out as Flammie from the 1993 Nintendo game "Secret of Mana." The annual gamer expo's big draw this year, attendees told the Herald, is the chance to sample the soon-to-be-released virtual reality headsets and mind-bending VR games that industry insiders predict will change the way video games are played forever. "It's kind of like a Woodstock for gamers," said Robert Khoo, president of Penny Arcade, the group behind the mega event. In addition to the wildly popular state-of-the-art virtual reality headsets, gamers stood in line for hours to play unreleased video games, test out the latest gadgets and compete in ultra-competitive tournaments.

Recently, there has been rising interest in Bayesian optimization -- the optimization of an unknown function with assumptions usually expressed by a Gaussian Process (GP) prior. We study an optimization strategy that directly uses an estimate of the argmax of the function. This strategy offers both practical and theoretical advantages: no tradeoff parameter needs to be selected, and, moreover, we establish close connections to the popular GP-UCB and GP-PI strategies. Our approach can be understood as automatically and adaptively trading off exploration and exploitation in GP-UCB and GP-PI. We illustrate the effects of this adaptive tuning via bounds on the regret as well as an extensive empirical evaluation on robotics and vision tasks, demonstrating the robustness of this strategy for a range of performance criteria.

This paper addresses an open challenge in educational data mining, i.e., the problem of automatically mapping online courses from different providers (universities, MOOCs, etc.) onto a universal space of concepts, and predicting latent prerequisite dependencies (directed links) among both concepts and courses. We propose a novel approach for inference within and across course-level and concept-level directed graphs. In the training phase, our system projects partially observed course-level prerequisite links onto directed concept-level links; in the testing phase, the induced concept-level links are used to infer the unknown course-level prerequisite links. Whereas courses may be specific to one institution, concepts are shared across different providers. The bi-directional mappings enable our system to perform interlingua-style transfer learning, e.g. treating the concept graph as the interlingua and transferring the prerequisite relations across universities via the interlingua. Experiments on our newly collected datasets of courses from MIT, Caltech, Princeton and CMU show promising results.

This paper considers the problem of matrix completion when some number of the columns are completely and arbitrarily corrupted, potentially by a malicious adversary. It is well-known that standard algorithms for matrix completion can return arbitrarily poor results, if even a single column is corrupted. One direct application comes from robust collaborative filtering. Here, some number of users are so-called manipulators who try to skew the predictions of the algorithm by calibrating their inputs to the system. In this paper, we develop an efficient algorithm for this problem based on a combination of a trimming procedure and a convex program that minimizes the nuclear norm and the $\ell_{1,2}$ norm. Our theoretical results show that given a vanishing fraction of observed entries, it is nevertheless possible to complete the underlying matrix even when the number of corrupted columns grows. Significantly, our results hold without any assumptions on the locations or values of the observed entries of the manipulated columns. Moreover, we show by an information-theoretic argument that our guarantees are nearly optimal in terms of the fraction of sampled entries on the authentic columns, the fraction of corrupted columns, and the rank of the underlying matrix. Our results therefore sharply characterize the tradeoffs between sample, robustness and rank in matrix completion.

Sum-Product Networks (SPNs) are a class of expressive yet tractable hierarchical graphical models. LearnSPN is a structure learning algorithm for SPNs that uses hierarchical co-clustering to simultaneously identifying similar entities and similar features. The original LearnSPN algorithm assumes that all the variables are discrete and there is no missing data. We introduce a practical, simplified version of LearnSPN, MiniSPN, that runs faster and can handle missing data and heterogeneous features common in real applications. We demonstrate the performance of MiniSPN on standard benchmark datasets and on two datasets from Google's Knowledge Graph exhibiting high missingness rates and a mix of discrete and continuous features.