Before it started scampering after the machine learning chip market in 2016, but after it was founded at the University of Michigan in 2012, Mythic was trying to build embedded chips that would let surveillance drones run software modeled after the human brain. Part of the funding for the company, then known as Isocline, came from the Department of Defense. But after relaunching two years ago, Mythic refocused on embedded devices like autonomous cars and security cameras. Now the company is only a few months from sampling chips based on an aggressively ambitious architecture, which uses analog computing inside flash memory cells to accelerate machine learning tasks like facial recognition. Helping it over the finish line is $40 million raised last month from new and existing investors, including SoftBank Ventures, Draper Fisher Jurvetson and Lux Capital.

A day after President Trump promised to slash the red tape involved in weapons sales, the administration announced on Thursday a new policy that could vastly expand sales of armed drones, a contentious emblem of the shift toward remotely controlled warfare. That change, in addition to a newly released update to the policy governing which nations are allowed to buy sophisticated American-made weapons, is intended to accelerate arms sales, a key priority of Mr. Trump. The president seemed to foreshadow the new policies on Wednesday night, when he said at a news conference with Prime Minister Shinzo Abe of Japan that after allies order weapons from the United States, "we will get it taken care of, and they will get their equipment rapidly." "It would be, in some cases, years before orders would take place because of bureaucracy with Department of Defense, State Department," Mr. Trump said. It's now going to be a matter of days.

The proposed changes include permitting IPOs that restrict shareholders' voting rights, secondary listings by Chinese and international companies already listed elsewhere and primary listings by unprofitable biotech firms. The reforms are set to become effective April 30. The exchange will begin taking listing applications in early May, Mr. Li said. "This probably is the largest reform we've ever had in the last 25 years," he said, adding that it's "only a matter of time" before the likes of Alibaba and Xiaomi list in Hong Kong. Mr. Li is one of several speakers who are discussing some of the most compelling ideas emerging globally.

Algorithmic defenses against adversarial examples remain an extremely open and challenging problem, with recent state-of-the-art defenses on ImageNet still achieving only 27.9% and 46.7% top-1 accuracy for white- and black-box PGD attacks, respectively, as of March 2018 \citepkannan2018adversarial. Unfortunately, despite the explosive emergence of defense strategies, there does not appear to be an easy algorithmic fix for the adversarial problem available in the short term. For example, one recent analysis investigated a series of promising methods that relied on gradient obfuscation, and demonstrated that they could be quickly broken \citepathalye2018obfuscated. Despite this, we also note that principled approaches to adversarial robustness are beginning to show promise. For example, several papers have demonstrated what appears to be both high accuracy and strong adversarial robustness on smaller datasets such as MNIST, \citepmadry2017towards,kannan2018adversarial, and there have also been several results including theoretical guarantees of adversarial robustness, albeit on small datasets and/or with still-insufficient accuracy \citepkolter2017provable, raghunathan2018certified, dvijotham2018dual.

IBM has launched a range of new machine learning-based security tools that it says can help firms identify and fight off cyber attacks. The company has unveiled its next-generation Resilient Incident Response Platform with'Intelligent Orchestration', which is aimed at helping enterprises respond to breaches more quickly and effectively. From a single platform, companies can access case incident management, orchestration, automation, intelligence, and "deep two-way partnership" features, said IBM in an announcement. Big Blue said its experts spent around 200,000 hours in research and development to create the software. In a bid to combine machine and human intelligence, the firm has also launched X-Force Threat Management Services, a new artificial intelligence-driven solution that automates threat detection.

Senator Maria Cantwell, D-Wash., just drafted forward-looking legislation that aims to establish a select committee of experts to advise agencies across the government on the economic impact of federal artificial intelligence. The move is an early step toward formalizing the exploration of AI in a government context. But it could ultimately contribute to jump-starting AI-focused programs that help stimulate the United States economy, benefit citizens, uphold data security and privacy, and eventually ensure America is successful during the initial introduction of this important technology to U.S. consumers. The presence of legislation could also lend legitimacy to the prospect of near-term government investment in AI innovation -- something that may even sway Treasury Secretary Steve Mnuchin and others away from their belief that the impact of AI won't be felt for years to come. Indeed, other than a few economic impact and policy reports conducted by the Obama Administration -- led by former U.S. Chief Data Scientist DJ Patil and other tech-minded government leaders -- this is the first policy effort toward moving the U.S. public sector past acknowledging its significance, and toward fully embracing AI technology.

To understand today's headlines about China's approach to issues such as trade, foreign policy or internet censorship, turn to its past. The country is perhaps more aware of its own history than any other major society on earth. That remembering is certainly partial - events like Mao's Cultural Revolution are still very difficult to discuss within China itself. But it is striking how many echoes of the past can be found in its present. China remembers a time when it was forced to trade against its will.

If you think about it, we are all part of a supply chain -- as producers, consumers or distributors, often playing multiple roles. It should not surprise anyone that Artificial Intelligence has a role here too. Now, supply chain management is nothing new, even if the term itself was coined relatively recently, in the 1980s. Supply chain management traces its roots to the early 20th century, when the development of the assembly line necessitated a new system to manage the flow of raw materials and finished goods. Although the science (and art) of supply chain management remains primarily concerned with this, the business landscape has evolved significantly. The increase in international trade, shift towards just-in-time production, and rise of e-commerce are just some of the factors that have changed how the game is played.

In this work, we focus on the task of generating natural language descriptions from a structured table of facts containing fields (such as nationality, occupation, etc) and values (such as Indian, actor, director, etc). One simple choice is to treat the table as a sequence of fields and values and then use a standard seq2seq model for this task. However, such a model is too generic and does not exploit task-specific characteristics. For example, while generating descriptions from a table, a human would attend to information at two levels: (i) the fields (macro level) and (ii) the values within the field (micro level). Further, a human would continue attending to a field for a few timesteps till all the information from that field has been rendered and then never return back to this field (because there is nothing left to say about it). To capture this behavior we use (i) a fused bifocal attention mechanism which exploits and combines this micro and macro level information and (ii) a gated orthogonalization mechanism which tries to ensure that a field is remembered for a few time steps and then forgotten. We experiment with a recently released dataset which contains fact tables about people and their corresponding one line biographical descriptions in English. In addition, we also introduce two similar datasets for French and German. Our experiments show that the proposed model gives 21% relative improvement over a recently proposed state of the art method and 10% relative improvement over basic seq2seq models. The code and the datasets developed as a part of this work are publicly available.

In recent years, it has been found that neural networks can be easily fooled by adversarial examples, which is a potential safety hazard in some safety-critical applications. Many researchers have proposed various method to make neural networks more robust to white-box adversarial attacks, but an effective method have not been found so far. In this short paper, we focus on the robustness of the features learned by neural networks. We show that the features learned by neural networks are not robust, and find that the robustness of the learned features is closely related to the resistance against adversarial examples of neural networks. We also find that adversarial training against fast gradients sign method (FGSM) does not make the leaned features very robust, even if it can make the trained networks very resistant to FGSM attack. Then we propose a method, which can be seen as an extension of adversarial training, to train neural networks to learn more robust features. We perform experiments on MNIST and CIFAR-10 to evaluate our method, and the experiment results show that this method greatly improves the robustness of the learned features and the resistance to adversarial attacks.