Deep neural networks have lately shown tremendous performance in various applications including vision and speech processing tasks. However, alongside their ability to perform these tasks with such high accuracy, it has been shown that they are highly susceptible to adversarial attacks: a small change of the input would cause the network to err with high confidence. This phenomenon exposes an inherent fault in these networks and their ability to generalize well. For this reason, providing robustness to adversarial attacks is an important challenge in networks training, which has led to an extensive research. In this work, we suggest a theoretically inspired novel approach to improve the networks' robustness. Our method applies regularization using the Frobenius norm of the Jacobian of the network, which is applied as post-processing, after regular training has finished. We demonstrate empirically that it leads to enhanced robustness results with a minimal change in the original network's accuracy.

Georgia Tech researchers are using deep learning to identify early signs of heart failure. In a paper published by the Journal of the American Medical Informatics Association (JAMIA), Georgia Tech's School of Computational Science and Engineering Associate Professor Jimeng Sun and Ph.D. student Edward Choi present a pioneering method for analyzing vast amounts of personal health record data that addresses temporality in the data – something previously ignored by conventional machine learning models in health care applications. The new research, funded by the National Institutes of Health in collaboration with Sutter Health, uses a deep learning model to enable earlier detection of the incidents and stages that often lead to heart failure within 6-18 months. To achieve this, Sun and Choi use a recurrent neural network (RNN) to model temporal relations among events in electronic health records. Temporal relationships communicate the ordering of events or states in time. This type of relation is traditionally used in natural language processing.

Most manufacturers have connected their operational technology – including industrial control systems and robotic equipment –to the internet, yet the lack of basic security protocols leave these companies open to cyberattacks. Dewan Chowdhury, founder of MalCrawler, said that many robotics that work as part of industrial systems on manufacturing floors are still leveraging outdated and unsupported operating systems – such as Windows XP. Chowdhury presented his research at a SAS session titled "Hack Your Robot". "Even before the robotics, the issue is that the programs that control the robotics are completely wide open to vulnerabilities," said Chowdhury. For manufacturing companies, cybersecurity threats are beginning to make headlines.

It's been a couple of years since AI-controlled bots fragged each other in an epic Doom deathmatch. Now, EA's Search for Extraordinary Experiences Division, or SEED, has taught self-learning AI agents to play Battlefield 1. Each character in the basic match uses a model based on neural-network training to learn how to play the game via trial and error. The AI-controlled troops in the game learned how to play after watching human players, then parallel training against other bots. The AI soldiers even learned how to pick up ammo or health when they're running low, much like you or I do.

After revelations that political consulting firm Cambridge Analytica allegedly appropriated Facebook user data to advise Donald Trump's 2016 U.S. presidential campaign, many are calling for greater regulation of social media networks, saying a'massive data breach' has occurred. The idea that governments can regulate their way into protecting citizen privacy is appealing, but I believe it misses the mark. What happened with Cambridge Analytica wasn't a breach or a leak. It was a wild violation of academic research ethics. The CEO finally broke his silence on the misuse of 51 million users' data Wednesday evening, outlining three steps the firm plans to take to prevent something like this from happening again.

Kathryn Hume and Carole Piovesan are powerful forces in the Toronto artificial intelligence community. Kathryn Hume is Vice President of product and strategy for Integrate.ai Carole Piovesan is a lawyer at McCarthy Tétrault LLP and the firm's lead in the area of artificial intelligence. Piovesan has appeared before various administrative tribunals, at all levels of court in Ontario, as well as at the Supreme Court of Canada. Hume and Piovesan are widely respected speakers and writers on AI and excel at communicating how AI and machine learning technologies work in everyday language.

The awful news is that one of Uber's self-driving cars hit and killed a pedestrian in Tempe, Arizona. If anything good is to come from the tragedy, cities need to seize this opportunity to change minds. Right now, while the companies running testbeds in American metropolises are forced to pause, city leaders have a chance to shape the future of autonomous vehicles and ensure they are part of holistic efforts to improve equity and quality of life for all residents. But if politicians simply introduce self-driving cars without conditions, we can expect tragedies on multiple levels. Susan Crawford (@scrawford) is an Ideas contributor for WIRED, a professor at Harvard Law School, author of Captive Audience: The Telecom Industry and Monopoly Power in the New Gilded Age, coauthor of The Responsive City, and a longtime columnist and blogger about tech policy.

This week, Raytheon announced it successfully tested its anti-drone technology. The advanced high-power microwave and laser dune buggy brought down 45 unmanned aerial vehicles (UAVs) and drones at a U.S. Army exercise that was held in Fort Sill, Oklahoma. The microwave system was able to bring down multiple UAVs at once when the devices swarmed, while the high energy laser (HEL) was able to identify and shoot down 12 Class I and II UAVs, as well as six different stationary devices that propelled mortar rounds. The equipment is intended to protect US troops against drones; it's self-contained and easy to deploy in a tense situation. The U.S. Air Force Research Laboratory worked with Raytheon to develop this counter-drone and UAV tech.

As policymakers debate the government's role in developing artificial intelligence, a House bill aims to shed light on the emerging technology's role in strengthening national security. The National Security Commission on Artificial Intelligence Act would create an independent panel to explore recent advancements in artificial intelligence and assess the economic and national security impacts of the budding technology. Introduced Tuesday by Rep. Elise Stefanik, R-N.Y., who heads the House Armed Services Subcommittee on Emerging Threats and Capabilities, the legislation would provide direction for agencies looking to steer the growth of AI in the coming years. With some experts calling artificial intelligence "the biggest economic and technological revolution" of our lifetimes, maintaining an edge in the field could prove critical to America's position on the world stage. "Artificial intelligence is a constantly developing technology that will likely touch every aspect of our lives," Stefanik said in a statement.

SANAA, Yemen – Yemeni tribal leaders say a suspected U.S. drone strike has killed seven alleged al-Qaida operatives in the central Marib province. They said Thursday's strike hit a house believed to have been used by the militants. The U.S. is believed to have carried out at least five drone strikes in Yemen since the beginning of March. The tribal leaders spoke on condition of anonymity for fear of reprisals. Al-Qaida in the Arabian Peninsula, as the Yemen affiliate is known, has long been seen as the global network's most dangerous branch.