Artificial Intelligence, or AI, has become a buzzword around the world. Just this Tuesday, Google's AI program AlphaGo defeated current Go world champion Ke Jie in the first out of three games in China. As AI technology emerges and AI-powered products become part of our lives, it makes people wonder its impact and to what extent people actually know about AI. With about 700 million users, China's news aggregator giant Toutiao conducted a survey in China with 3,088 valid samples, and released a report on how its users perceive AI. In short, the report shows that people in China have a vague understanding of AI products.

You enter your password incorrectly too many times and get locked out of your account; your colleague sets up access to her work email on a new device; someone in your company clicks on an emailed "Google Doc" that is actually a phishing link -- initially thought to be how the recent spread of the WannaCry computer worm began. Each of these events leaves a trace in the form of information flowing through a computer network. But which ones should the security systems protecting your business against cyber attacks pay attention to and which should they ignore? And how do analysts tell the difference in a world that is awash with digital information? The answer could lie in human researchers tapping into artificial intelligence and machine learning, harnessing both the cognitive power of the human mind and the tireless capacity of a machine.

Researchers have developed a waltzing robot that can teach people how to dance. In contrast to previous dancing robot, this one can take the lead, allowing the robot to teach dance sequences. While the system has been developed for dancing, it could also have other applications including physical rehabilitation and sports training. According to the researchers, experiments with multiple volunteers showed safe interaction guided by the robot and improvement in the volunteers' dancing abilities The system adjusts its difficulty mode based on the user's number of previous practices and performance history. The bot, which stands 1.8 meters tall (5 feet 9 inches), was designed by researchers at Tohoku University in Japan. According to the authors of the study, the bot its designed for contact with adults with heights ranging from 1.5 meters (4 feet 9 inches) to 1.9 meters (6 feet two inches) meters tall.

We empirically characterize the performance of discriminative and generative LSTM models for text classification. We find that although RNN-based generative models are more powerful than their bag-of-words ancestors (e.g., they account for conditional dependencies across words in a document), they have higher asymptotic error rates than discriminatively trained RNN models. However we also find that generative models approach their asymptotic error rate more rapidly than their discriminative counterparts---the same pattern that Ng & Jordan (2001) proved holds for linear classification models that make more naive conditional independence assumptions. Building on this finding, we hypothesize that RNN-based generative classification models will be more robust to shifts in the data distribution. This hypothesis is confirmed in a series of experiments in zero-shot and continual learning settings that show that generative models substantially outperform discriminative models.

Dubai Police have revealed their first robot officer, giving it the task of patrolling the city's malls and tourist attractions. People will be able to use it to report crimes, pay fines and get information by tapping a touchscreen on its chest.

We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry. Four years ago, a paper by Yann LeCun and his collaborators was rejected by the leading computer vision conference on the grounds that it used neural networks and therefore provided no insight into how to design a vision system. At the time, most computer vision researchers believed that a vision system needed to be carefully hand-designed using a detailed understanding of the nature of the task. They assumed that the task of classifying objects in natural images would never be solved by simply presenting examples of images and the names of the objects they contained to a neural network that acquired all of its knowledge from this training data. What many in the vision research community failed to appreciate was that methods that require careful hand-engineering by a programmer who understands the domain do not scale as well as methods that replace the programmer with a powerful general-purpose learning procedure.

Stanford University visiting researcher Alireza Marandi (right) and post-doctoral scholar Peter McMahon inspect a prototype of a new light-based computer. A 20th-century theoretical model of the way magnetism develops in cooling solids is driving the development of analog computers that could deliver results with much less electrical power than today's super-computers. But the work may instead yield improved digital algorithms rather than a mainstream analog architecture. Helmut Katzgraber, associate professor at Texas A&M in College Station, TX, argues, "There is a deep synergy between classical optimization, statistical physics, high-performance computing, and quantum computing. Those things really go hand in hand. Nature is the best optimizer out there. Lightning typically chooses the path of least resistance. A soap bubble will always give you the minimal surface."

The history of battle knows no bounds, with weapons of destruction evolving from prehistoric clubs, axes, and spears to bombs, drones, missiles, landmines, and systems used in biological and nuclear warfare. More recently, lethal autonomous weapon systems (LAWS) powered by artificial intelligence (AI) have begun to surface, raising ethical issues about the use of AI and causing disagreement on whether such weapons should be banned in line with international humanitarian laws under the Geneva Convention. Much of the disagreement around LAWS is based on where the line should be drawn between weapons with limited human control and autonomous weapons, and differences of opinion on whether more or less people will lose their lives as a result of the implementation of LAWS. There are also contrary views on whether autonomous weapons are already in play on the battlefield. Ronald Arkin, Regents' Professor and Director of the Mobile Robot Laboratory in the College of Computing at Georgia Institute of Technology, says limited autonomy is already present in weapon systems such as the U.S. Navy's Phalanx Close-In Weapons System, which is designed to identify and fire at incoming missiles or threatening aircraft, and Israel's Harpy system, a fire-and-forget weapon designed to detect, attack, and destroy radar emitters.

The city wants 25 percent of its police force to be robots by 2030.

The Donald Trump administration has asked Congress to give the federal government the ability to track and destroy any type of drone flying on domestic soil, a document obtained by the New York Times reveals. Under the proposal, government agencies and law enforcement would have the ability to monitor and take action against any unmanned aircraft system flying over an area designated for protection. The draft legislation would authorize government agencies to track, take control of and destroy any drone that it determines to be a threat to a "covered facility, location, or installation," which could refer to any number of locations. The proposal would call for the government to respect "privacy, civil rights and civil liberties" when exercising its power to take down drones, but courts would be given no jurisdiction to hear lawsuits filed by drone operators who have their vehicles downed. An exception for drones would be created through the proposal in U.S. hacking and surveillance laws and in Federal Aviation Administration (FAA) aircraft regulations, which currently protect the unmanned aircraft.