That people would generally prefer to minimize casualties in a hypothetical autonomous car crash has been found to be true in past research, but what happens when people are presented with more complex scenarios? And what happens when autonomous vehicles must choose between two scenarios in which at least one individual could die? Who might those vehicles save, and on what basis do they make those ethical judgments? It may sound like a nightmarish spin on "would you rather," but researchers say such thought experiments are necessary to the programming for autonomous vehicles and the policies that regulate them. What's more, the responses around these difficult ultimatums may vary across cultures, revealing there's no universality in what people believe to be a morally superior option.

Liquid metal robots that can change their form and repair from damage just like the androids of the Terminator films could soon become a reality. Researchers in China have developed a palm-sized prototype inspired by T-1000 from the science fiction franchise, albeit a lot less sinister. The small, shape-shifting robot could be used to access environments that would be difficult for a human or fixed-shape bot to navigate, such as disaster zones. Liquid metal robots that can change their form and repair from damage just like the androids of the Terminator films could soon become a reality. The prototype, created by a team from the University of Science and Technology of China and the University of Wollongong in Australia is made up of a small plastic wheel, a lithium battery, and drops of gallium, a soft silvery metal, according to the South China Morning Post.

We present a novel introspective variational autoencoder (IntroVAE) model for synthesizing high-resolution photographic images. IntroVAE is capable of self-evaluating the quality of its generated samples and improving itself accordingly. Its inference and generator models are jointly trained in an introspective way. On one hand, the generator is required to reconstruct the input images from the noisy outputs of the inference model as normal VAEs. On the other hand, the inference model is encouraged to classify between the generated and real samples while the generator tries to fool it as GANs. These two famous generative frameworks are integrated in a simple yet efficient single-stream architecture that can be trained in a single stage. IntroVAE preserves the advantages of VAEs, such as stable training and nice latent manifold. Unlike most other hybrid models of VAEs and GANs, IntroVAE requires no extra discriminators, because the inference model itself serves as a discriminator to distinguish between the generated and real samples. Experiments demonstrate that our method produces high-resolution photo-realistic images (e.g., CELEBA images at \(1024^{2}\)), which are comparable to or better than the state-of-the-art GANs.

It is widely believed that learning good representations is one of the main reasons for the success of deep neural networks. Although highly intuitive, there is a lack of theory and systematic approach quantitatively characterizing what representations do deep neural networks learn. In this work, we move a tiny step towards a theory and better understanding of the representations. Specifically, we study a simpler problem: How similar are the representations learned by two networks with identical architecture but trained from different initializations. We develop a rigorous theory based on the neuron activation subspace match model. The theory gives a complete characterization of the structure of neuron activation subspace matches, where the core concepts are maximum match and simple match which describe the overall and the finest similarity between sets of neurons in two networks respectively. We also propose efficient algorithms to find the maximum match and simple matches. Finally, we conduct extensive experiments using our algorithms. Experimental results suggest that, surprisingly, representations learned by the same convolutional layers of networks trained from different initializations are not as similar as prevalently expected, at least in terms of subspace match.

In this article, we introduce a new mode for training Generative Adversarial Networks (GANs). Rather than minimizing the distance of evidence distribution $\tilde{p}(x)$ and the generative distribution $q(x)$, we minimize the distance of $\tilde{p}(x_r)q(x_f)$ and $\tilde{p}(x_f)q(x_r)$. This adversarial pattern can be interpreted as a Turing test in GANs. It allows us to use information of real samples during training generator and accelerates the whole training procedure. We even find that just proportionally increasing the size of discriminator and generator, it succeeds on 256x256 resolution without adjusting hyperparameters carefully.

In what ways is artificial intelligence leveraged to harm people? Currently most AI is used for purposes that are either beneficial to people, or not actively harmful. But any powerful tool can be abused, and we should expect increasing weaponization of AI. Social manipulation: Companies and organizations use big data to spread propaganda, influence elections and foment social discord via targeted marketing. Political actors like Cambridge Analytica and the Internet Research Agency have been prominent in this.

From virtual reality police training programmes to gun-toting drones and iris scanners, a public security expo in China showed the range of increasingly high-tech tools available to the country's police. The exhibition, which ran Tuesday to Friday in Beijing, emphasised surveillance and monitoring technology just as the Communist government's domestic security spending has skyrocketed. Facial-recognition screens analysing candid shots of conference attendees were scattered around the exhibition hall, while other vendors packed their booths with security cameras. From virtual reality police training programmes to gun-toting drones and iris scanners, a public security expo in China showed the range of increasingly high-tech tools available to the country's police. More innocuous applications, like smart locks for homes and big data applications to reduce traffic congestion, also occupied large swathes of the conference.

Since the arrival of the fall armyworm (Spodoptera frugiperda) caterpillar in West Africa in early 2016, true to its name, it has been marching quickly and mercilessly through the continent, eating maize (corn) along with sorghum, millet, and rice and causing billions of dollars in crop losses. It has now been confirmed or reported in every sub-Saharan African country and was recently found in southern India, beginning its likely spread into much of the Asian continent. The time for eradication has long passed, and scientists, NGOs, and governments are now focused on control. For some, this means chemical pesticides, but these are expensive and many smallholders do not know how to safely apply the chemicals, making them a threat to human and environmental health, including the survival of other insects and their predators. Additionally, many farmers have said that even when they spray the pesticides, they are ineffective.

Artificial intelligence (AI) that reasons like a human remains elusive, but Honda hopes to make inroads. The Tokyo company's U.S.-based Research Institute today announced a collaboration with three academic institutions -- the Massachusetts Institute of Technology (MIT), the University of Pennsylvania (Penn), and the University of Washington -- to advance the field of artificial cognition. MIT's Computer Science and Artificial Intelligence (CSAIL) lab, in partnership with Penn's School of Engineering and Applied Science and the University of Washington's Paul G. Allen School of Computer Science & Engineering, will develop prototypes, working examples, and demonstrations of what Honda calls the "mechanisms of curiosity." Specifically, MIT CSAIL will focus its efforts on systems capable of predicting future percepts -- concepts developed as a consequence of perception -- and the effect of future actions, while Penn's engineering department and the Paul G. Allen School will develop perception models informed by biology and robots that can work safely in human environments. Grants will fund the first leg of research.

Cities in China are growing rapidly in terms of both size and complexity. Governments have been searching for new technologies to make cities more efficient, and smart mobility has been the top priority in all solutions. The past few years have seen a paradigm shift for smart mobility in China, that is, data-centric companies, mostly Internet companies, are taking a leading role in such initiatives instead of governments and academic researchers. For example, Alibaba, Baidu, Tencent, Ctrip, and Didi, among others, are spearheading the smart mobility initiatives. The driving force is twofold: these companies have accumulated a huge volume of data and invested a great deal of resources in the AI arena.