More than half a million people are diagnosed with cancers of the head and neck each year, many of whom choose to undergo radiotherapy. But it's a delicate process: The surrounding tissue can be severely damaged if it isn't carefully isolated prior to treatments. In partnership with the University College London Hospital, Google subsidiary DeepMind is exploring ways artificial intelligence (AI) can aid in the segmentation process. It today announced a significant step forward in the pursuit of that vision: validation of a model that exhibits "near-human performance" on CT scans. "Automated … segmentation has the potential to address these challenges but, to date, performance of available solutions in clinical practice has proven inferior to that of expert human operators," the researchers wrote.

A few years after the Great Recession, you couldn't scroll through Google Reader without seeing the word "disrupt." TechCrunch named a conference after it, the New York Times named a column after it, investor Marc Andreessen warned that "software disruption" would eat the world; not long after, Peter Thiel, his fellow Facebook board member, called "disrupt" one of his favorite words. The term "disruptive innovation" was coined by Harvard Business School professor Clayton Christensen in the mid-90's to describe a particular business phenomenon, whereby established companies focus on high-priced products for their existing customers, while disruptors develop simpler, cheaper innovations, introduce the products to a new audience, and eventually displace incumbents. PCs disrupted mainframes, discount stores disrupted department stores, cellphones disrupted landlines, you get the idea. In Silicon Valley's telling, however, "disruption" became shorthand for something closer to techno-darwinism.

Pedestrians walk under a surveillance camera, which is part of a facial recognition technology test in Berlin, Germany. Machine learning shines at tasks like this because it can recognize patterns and predict threats in massive data sets, all at machine speed. There is a temptation to hail major advances in technology as cure-alls for the challenges facing organizations and society today. The fanfare usually ends in disappointment, as the latest superhero technology doesn't live up to its expectations. Not surprisingly, machine learning, a domain within the broader field of artificial intelligence, has been hailed as the current be-all end-all answer in cybersecurity. As a result, it is currently at the peak of inflated expectations in Gartner's most recent Hype Cycle for Emerging Technologies.

When I moved to China in 2014 to start my company, the country was still in the slow process of opening up its massive economy and developing its major foreign policy initiative, One Belt One Road. This aims to open up the economy by reviving the old Silk Road trade route, connecting Europe and China by land and sea, at a cost of $4-8 trillion. The impact of China's opening and development can be felt in every industry, especially the start-up and venture capital sectors. In 2014, only two of the world's top 20 internet companies were from China. Today, that number has jumped to nine.

The Massachusetts Institute of Technology announced this week that it would invest $1 billion in a new college of computer engineering: the Stephen A. Schwarzman College of Computing. And when the new building hosts its first classes in 2022, it'll be the largest structural addition to MIT's campus since the 1950s. Yet another AI-forward institution of note is Carnegie Mellon University (CMU), which partnered with Bosch's Center for Artificial Intelligence on an $8 million research project that goes through 2023. CMU has the additional distinction of being the first university to offer an undergraduate degree in AI, and it neighbors the ARM Institute, a $250 million initiative focused on accelerating the advancement of transformative robotics technologies and education in the U.S. manufacturing industry. This week, I participated in a tour of startups in Pittsburgh's blossoming robotics and automation industry, the majority of which draw on CMU not just for funding, but for expertise.

But Väärsi's assurance isn't likely to soothe those who fear that armed robots are a Pandora's box that will bite their makers, especially as advanced AI enables machines to function more and more autonomously. Israel and South Korea already have armed robots patrolling their borders, while Russia has tested its Uran-9--a mini-tank armed with a 30-millimeter cannon--in Syria. Even the United States is getting into the armed robot game as it develops kits that can transform a normal vehicle into an autonomous one. The U.S. army already has an armed M113 armored personnel carrier as a remote-controlled test vehicle.

Let's face it, Boston Dynamics is the SpaceX of Robotics, and if you're a fan of robotics like I am, you can't wait to see what they come up with next. While we wait, let's check out some robots developed in Japan capable of doing amazing things. First is the HRP-5P Developed by Japan's National Institute of Advanced Industrial Science or AIST and Technology. There is legitimate concern over robots taking away jobs in the future, Japan, however, wants this to happen. This is because the country will have a workforce shortage in the future due to declining birth rates and strict immigration laws.

There are many examples of 'wisdom of the crowd' effects in which the large number of participants imparts confidence in the collective judgment of the crowd. But how do we form an aggregated judgment when the size of the crowd is limited? Whose judgments do we include, and whose do we accord the most weight? This paper considers this problem in the context of geopolitical event forecasting, where volunteer analysts are queried to give their expertise, confidence, and predictions about the outcome of an event. We develop a forecast aggregation model that integrates topical information about a question, meta-data about a pair of forecasters, and their predictions in a deep siamese neural network that decides which forecasters' predictions are more likely to be close to the correct response. A ranking of the forecasters is induced from a tournament of pair-wise forecaster comparisons, with the ranking used to create an aggregate forecast. Preliminary results find the aggregate prediction of the best forecasters ranked by our deep siamese network model consistently beats typical aggregation techniques by Brier score.

Deep neural networks (DNNs) are computationally/memory-intensive and vulnerable to adversarial attacks, making them prohibitive in some real-world applications. By converting dense models into sparse ones, pruning appears to be a promising solution to reducing the computation/memory cost. This paper studies classification models, especially DNN-based ones, to demonstrate that there exists intrinsic relationships between their sparsity and adversarial robustness. Our analyses reveal, both theoretically and empirically, that nonlinear DNN-based classifiers behave differently under $l_2$ attacks from some linear ones. We further demonstrate that an appropriately higher model sparsity implies better robustness of nonlinear DNNs, whereas over-sparsified models can be more difficult to resist adversarial examples.

Abstract--In the era of "big data", analysts usually explore various statistical models or machine learning methods for observed data in order to facilitate scientific discoveries or gain predictive power. Whatever data and fitting procedures are employed, a crucial step is to select the most appropriate model or method from a set of candidates. Model selection is a key ingredient in data analysis for reliable and reproducible statistical inference or prediction, and thus central to scientific studies in fields such as ecology, economics, engineering, finance, political science, biology, and epidemiology. There has been a long history of model selection techniques that arise from researches in statistics, information theory, and signal processing. A considerable number of methods have been proposed, following different philosophies and exhibiting varying performances. The purpose of this article is to bring a comprehensive overview of them, in terms of their motivation, large sample performance, and applicability. We provide integrated and practically relevant discussions on theoretical properties of state-ofthe-art model selection approaches. We also share our thoughts on some controversial views on the practice of model selection. Vast development in hardware storage, precision instrument manufacture, economic globalization, etc. have generated huge volumes of data that can be analyzed to extract useful information. Typical statistical inference or machine learning procedures learn from and make predictions on data by fitting parametric or nonparametric models (in a broad sense). However, there exists no model that is universally suitable for any data and goal. This research was funded in part by the Defense Advanced Research Projects Agency (DARPA) under grant number W911NF-18-1-0134. J. Ding and Y. Yang are with the School of Statistics, University of Minnesota, Minneapolis, Minnesota 55455, United States. V. Tarokh is with the Department of Electrical and Computer Engineering, Duke University, Durham, North Carolina 27708, United States. Therefore, a crucial step in a typical data analysis is to consider a set of candidate models (referred to as the model class), and then select the most appropriate one. In other words, model selection is the task of selecting a statistical model from a model class, given a set of data. There have been many overview papers on model selection scattered in the communities of signal processing [1], statistics [2], machine learning [3], epidemiology [4], chemometrics [5], ecology and evolution [6]. Despite the abundant literature on model selection, existing overviews usually focus on derivations, descriptions, or applications of particular model selection principles.