Competitors at this year's World Human Powered Speed Challenge are going to have to contend with this--a bullet-shaped bike designed by an artificially intelligent software program. In 2012, a bicycle screamed across a flat, open road of the Nevada Desert at an astounding 88.13 miles per hour, or 133.78 km/hr. This record, established by a Dutch team at the annual World Human Powered Speed Challenge, could now be in danger, owing to a new bike designed by researchers at IUT Annecy, with the help of computer scientists at Neural Concept, a Subsidiary of the Swiss Federal Institute of Technology in Lausanne (EPFL). To be fair, the IUT Annecy researchers can't take full credit for the bike's sleek, aerodynamic shape, nor can any human for that matter. You see, this machine was, in part, designed by another machine--an artificially intelligent program developed by researchers at Neural Concept, who are presenting their findings today in Stockholm, Sweden, at the International Conference on Machine Learning.

It may have a history stretching back 3,000 years, but Beijing is fast becoming today's modern'it' city. With amazingly fast internet, access to cutting-edge technology like facial recognition software, significant investment in artificial intelligence and an unrivalled cosmopolitan energy, China's capital is among the most exciting cities for enterprising expats. "'If I can make it there, I'll make it anywhere'," said German expat Clemens Sehi, referencing Frank Sinatra's ode to New York City. "If Sinatra lived today, he would probably sing about a city like Beijing." Sehi, who is creative director at Travellers Archive, says living in the city means you feel like you are "living in the new age" and always up to date.

Ready or not, autonomous robots are leaving laboratories to be tested in real-world contexts. With more and more people living in cities, these technologies offer ways to cope with ageing populations and poorly maintained infrastructures, while promoting safer transport, productive manufacturing and secure energy supplies. Urban "living labs" are one way scientists are trying to understand how autonomous robots – or Robotics and Autonomous Systems (RAS), to give them their full title – will affect our everyday lives. Autonomous robots are interconnected, interactive, cognitive and physical tools, which can perceive their environments, reason about events, make or revise plans and control their own actions. These technologies are designed to draw on big data and connect with the Internet of Things, to make our lives easier by increasing accuracy and efficiency.

Anastasia Gnezditskaia is a writer / analyst covering France, Benelux and Morocco. Based in Antwerp, Belgium, she has a background working for trade publications covering markets and their regulation in Washington, D.C., where she lived for 10 years. Following this she managed international development projects in Africa at the World Bank, and worked as a journalist covering Congress, federal government agencies and financial markets, including energy futures.

The idea of Artificial Intelligence has been around since the early 1900s, originally in the form of fictional writing and later seen in films. The minds of these writers and film producers imagined a world where the role of robots in society was elevated from the role of machines in their present society. These individuals imagined technological advances, which would provide a machine with the ability to process sets of information and make a decision based on the information that the machine was taught to process. Although exploration within the field of AI and robotic process automation evolved at a slow pace originally, advancement within this field is currently increasing at an exponential rate. Many ideas that were once considered merely dreams are becoming reality.

NVIDIA teamed up with researchers from Finland's Aalto University and MIT to teach an old AI a new trick. Their neural network can now fix grainy or pixelated images in your photo library just by looking at them. AIs have been able to do similar work for a while, but typically it required both a so-called noisy image (grainy, pixelated) and a noise-free one in order for the AI to learn how to make up the difference and clean up the photo. This new method, which is being presented at the International Conference on Machine Learning in Stockholm this week (assuming my invitation was lost in the mail?), no longer requires a noise-free image for the AI to remove artifacts, noise, grain, and automatically enhance your photos. Using deep-learning work, the AI can look at those so-called noisy images and make them clear even without looking at a clean image first.

Machine learning algorithms can help scientists predict chemical toxicity to a similar degree of accuracy as animal testing, according to a paper published this week in Toxicological Sciences. A whopping €3bn (over $3.5bn) is spent every year to study how the negative impacts of chemicals on animals like rats, rabbits or monkeys. The top nine most frequently tested safety experiments resulted in the death of the poor critters 57 per cent of the time in Europe in 2011. By using software, chemists may be able to spend less on animal testing and save more creatures. To demonstrate this, first, a team of researchers scoured through a range of databases to label 80,908 different chemicals.

THE GIST: Although media headlines imply we are already living in a future where AI has infiltrated every aspect of society, this actually sets unrealistic expectations about what AI can really do for humanity. Governments around the world are racing to pledge support to AI initiatives, but they tend to understate the complexity around deploying advanced machine learning systems in the real world. This article reflects on the risks of "AI solutionism": the increasingly popular belief that, given enough data, machine learning algorithms can solve all of humanity's problems. There is no AI solution for everything. All solutions come at a cost and not everything that can be automated should be.

Many deep models have been recently proposed for anomaly detection. This paper presents comparison of selected generative deep models and classical anomaly detection methods on an extensive number of non--image benchmark datasets. We provide statistical comparison of the selected models, in many configurations, architectures and hyperparamaters. We arrive to conclusion that performance of the generative models is determined by the process of selection of their hyperparameters. Specifically, performance of the deep generative models deteriorates with decreasing amount of anomalous samples used in hyperparameter selection. In practical scenarios of anomaly detection, none of the deep generative models systematically outperforms the kNN.

Classifying the complexity of arithmetic computations is a crucial endeavour in theoretical computer science. Particularly interesting are the decidability and complexity issues pertaining to iterative arithmetic computations, i.e. computations consisting of a repeated application of some set of arithmetic operations on some initial value. Examples of such problems include matrix powering over various semirings [16, 9], the Skolem problem ("Does a given linear recurrent sequence contain a zero term?") and its variants [26, 27, 6], or the related Orbit problem [21, 4]. It is often natural to consider bounded or finite-horizon variants of the problems, which ask, given a time horizon H, whether a certain property holds within the first H iterations of the computation [9]. In this paper, we study the complexity of finite-horizon arithmetic computations arising in algorithmic decision making and operations research.