If European Central Bank chief Mario Draghi appears slightly more downbeat at his regular news conference than before, it could foreshadow a possible move by the bank to trim its monetary policy stimulus. That's the conclusion of two Japanese researchers who have used artificial intelligence software to analyze split-second changes in Draghi's facial expressions at his news conferences following policy meetings. The findings follow a similar analysis by the same researchers of Draghi's Japanese counterpart, Haruhiko Kuroda, last year, which claimed to have identified a correlation between patterns in his facial expressions and subsequent policy changes. Yoshiyuki Suimon and Daichi Isami, the paper's authors, think that subtle changes in Draghi's facial expressions could reflect a sense of frustration Draghi might have been feeling before making policy adjustments. Their study covered Draghi's news conference from June 2016 to December 2017 and found signs of "sadness" preceding two recent major policy changes -- when the central bank announced a dovish tapering in December 2016 and another quantitative easing cutback in October last year.

WASHINGTON – President Donald Trump will soon make it easier to export some types of lethal U.S.-made drones to potentially dozens more allies and partners -- including Japan -- according to people familiar with the plan. Trump is expected to ease rules for such foreign sales under a long-delayed new policy on unmanned military aircraft due to be rolled out as early as this month, the first phase of a broader overhaul of arms export regulations. U.S. drone manufacturers, facing growing competition overseas especially from Chinese and Israeli rivals who often sell under lighter restrictions, have lobbied hard for the rule changes. The White House is expected to tout the move as part of Trump's "Buy American" initiative to create jobs and reduce the U.S. trade deficit. Human rights and arms control advocates, however, warn it risks fueling violence and instability in regions such as the Middle East and South Asia.

I generally expect better from the American Interest than this article. The fact of the matter is we've heard this argument re-emerge every few decades and it's always been wrong. Our modern fear that robots/AI/automation will steal all the jobs fits a classic script. Nearly 500 years ago, Queen Elizabeth I cited the same fear when she denied an English inventor named William Lee a patent for an automated knitting contraption. Did you know that Keynes predicted that automation would result in a 15 hours/week work week back in the 1930s?

Artificial Intelligence (AI) is once again a promising technology. The last time this happened was in the 1980s, and before that, the late 1950s through the early 1960s. In between, commentators often described AI as having fallen into "Winter," a period of decline, pessimism, and low funding. Understanding the field's more than six decades of history is difficult because most of our narratives about it have been written by AI insiders and developers themselves, most often from a narrowly American perspective.[1] In addition, the trials and errors of the early years are scarcely discussed in light of the current hype around AI, heightening the risk that past mistakes will be repeated. How can we make better sense of AI's history and what might it tell us about the present moment?

Uber is confirming that one of its self-driving vehicles struck and killed a pedestrian in the Phoenix metro area Sunday night. Company officials say Uber is halting all of its self-driving testing as of Monday as the investigation continues. A handout photo from Uber shows one of its Volvo self-driving SUVs in a desert setting. One of the company's vehicles struck and killed a pedestrian in Tempe, Ariz., Sunday night. PHOENIX -- The operator behind the wheel of a self-driving Uber vehicle that hit and killed a 49-year-old woman in Tempe Sunday night had served almost four years in an Arizona prison in the early 2000s on an attempted armed robbery conviction.

The possibility of creating sentient machines that can think and act like humans raises many ethical issues. We're already encountering reinforced human bias in AI algorithms and with autonomous "killer" robots looming on the horizon, an open discussion on the perils of unchecked AI is even more imperative. In celebration of Women's History Month, we've highlighted 12 brilliant women leading this much-needed discussion on AI & ethics and development of responsible AI solutions that will benefit everyone. Let me know of any others we should highlight in the comments below or tweet @MiaD #TiEInflect. First person on our list is Joy Buolamwini, founder of Algorithmic Justice League to fight bias in Machine Learning.

Uber is confirming that one of its self-driving vehicles struck and killed a pedestrian in the Phoenix metro area Sunday night. Company officials say Uber is halting all of its self-driving testing as of Monday as the investigation continues. A handout photo from Uber shows one of its Volvo self-driving SUVs in a desert setting. One of the company's vehicles struck and killed a pedestrian in Tempe, Ariz., Sunday night. The death of an Arizona pedestrian by a self-driving car has resurfaced concerns that this futuristic technology is too risky to test in public places.

The User Interfaces and Scheduling and Planning (UISP) Workshop had its inaugural meeting at the 2017 International Conference on Automated Scheduling and Planning (ICAPS). The UISP community focuses on bridging the gap between automated planning and scheduling technologies and user interface (UI) technologies. Planning and scheduling systems need UIs, and UIs can be designed and built using planning and scheduling systems. The workshop participants included representatives from government organizations, industry, and academia with various insights and novel challenges. We summarize the discussions from the workshop as well as outline challenges related to this area of research, introducing the now formally established field to the broader user experience and artificial intelligence communities.

Internet of Things (IoT) technologies have made considerable recent advances in commercial applications, prompting new research on their use in military applications. Towards the development of an Internet of Battlefield Things (IoBT), capable of leveraging mixed commercial and military technologies, several unique challenges of the tactical environment present themselves. These challenges include development of methods for: (I) quickly gathering training data reflecting unforeseen learning/classification tasks; (II) incrementally learning over real-time data streams; (III) management of limited network bandwidth and connectivity between IoBT assets in data gathering and classification tasks. This paper provides a survey over classical and modern statistical learning theory, and how numerical optimization can be used to solve corresponding mathematical problems. The objective of this paper is to encourage the IoT and machine learning research communities to revisit the underlying mathematical underpinnings of stream-based learning, as applicable to IoBT-based systems.

Adversarial attack methods have demonstrated the fragility of deep neural networks. Their imperceptible perturbations are frequently able fool classifiers into potentially dangerous misclassifications. We propose a novel way to interpret adversarial perturbations in terms of the effective input signal that classifiers actually use. Based on this, we apply specially trained autoencoders, referred to as S2SNets, as defense mechanism. They follow a two-stage training scheme: first unsupervised, followed by a fine-tuning of the decoder, using gradients from an existing classifier. S2SNets induce a shift in the distribution of gradients propagated through them, stripping them from class-dependent signal. We analyze their robustness against several white-box and gray-box scenarios on the large ImageNet dataset. Our approach reaches comparable resilience in white-box attack scenarios as other state-of-the-art defenses in gray-box scenarios. We further analyze the relationships of AlexNet, VGG 16, ResNet 50 and Inception v3 in adversarial space, and found that VGG 16 is the easiest to fool, while perturbations from ResNet 50 are the most transferable.