The world's first ever underwater Artificial Intelligence colony will be created on the South China sea bed, Chinese President Xi Jinping claims. The base has been described as a'deep sea Atlantis' and will be used for unmanned submarine science and defence operations. Chinese officials and scientists familiar with the plans say that the deep sea station will analyse samples from the sea bed and send reports to the surface. Xi urged the scientists and engineers to'dare to do something that has never been done before' on a recent visit to the deep sea research institute in Hainan Province. China's unmanned submarine vehicle Qianlong III, pictured, could help to drive a subsea exploration programme and herald the arrival of an AI colony on the South China Sea bed, Chinese scientists and officials say'There is no road in the deep sea, we do not need to chase after other countries, we are the road,' President Xi said.

'My lens cover isn't off yet, but I just had to show you a first look at my new home.' While NASA has numerous Mars landings under its belt, similar attempts have proved a difficult hurdle for many missions. The Soviet Union never managed to land on Mars, and both attempts by the European Space Agency flopped. By contrast, just one of Nasa's previous eight attempts have failed. The new InSight lander has landed in a region known as Elysium Planitia.

The Bank of England has released a list of scientists who have been nominated to feature on the new £50 note. On the list are computing pioneers Alan Turing and Ada Lovelace, telephone inventor Alexander Graham Bell and astronomer Patrick Moore. The Bank received 174,112 nominations, of which 114,000 met the eligibility criteria. To be on the list, the individual must be real, deceased and have contributed to the field of science in the UK. The list, which includes more than 600 men and almost 200 women, includes black holes expert Stephen Hawking, penicillin discoverer Alexander Fleming, father of modern epidemiology John Snow, naturalist and zookeeper Gerald Durrell, fossil pioneer Mary Anning, British-Jamaican business woman and nursing pioneer Mary Seacole and Margaret Thatcher, who was a scientist before becoming prime minister.

The US space agency Nasa has landed a new robot on Mars after a dramatic seven-minute plunge to the surface of the Red Planet. The InSight probe aims to study the world's deep interior, and make it the only planet - apart from Earth - that has been examined in this way. Confirmation of touchdown came through on cue at 19:53 GMT. It ended an anxious wait in which the robot radioed home a series of updates on its descent. Nasa's mission control at California's Jet Propulsion Laboratory (JPL) erupted into cheers when it became clear InSight was safe on the ground.

Nasa's InSight spacecraft is about to endure "seven minutes of terror" and might end them dead on the Martian surface. The lander is the first spacecraft built to explore deep in the interior of Mars, digging down and trying to understand its core. But based on previous performance it may never get the chance. More Martian craft have got lost trying to make their way through the thin Martian atmosphere than have successfully landed. Even in recent years, spacecraft have regularly failed on their approach and smashed into the surface, not to be heard from again.

Recently, techniques for applying convolutional neural networks to graph-structured data have emerged. Graph convolutional neural networks (GCNNs) have been used to address node and graph classification and matrix completion. Although the performance has been impressive, the current implementations have limited capability to incorporate uncertainty in the graph structure. Almost all GCNNs process a graph as though it is a ground-truth depiction of the relationship between nodes, but often the graphs employed in applications are themselves derived from noisy data or modelling assumptions. Spurious edges may be included; other edges may be missing between nodes that have very strong relationships. In this paper we adopt a Bayesian approach, viewing the observed graph as a realization from a parametric family of random graphs. We then target inference of the joint posterior of the random graph parameters and the node (or graph) labels. We present the Bayesian GCNN framework and develop an iterative learning procedure for the case of assortative mixed-membership stochastic block models. We present the results of experiments that demonstrate that the Bayesian formulation can provide better performance when there are very few labels available during the training process.

In this work, we generalize semi-supervised generative adversarial networks (GANs) from classification problems to regression problems. In the last few years, the importance of improving the training of neural networks using semi-supervised training has been demonstrated for classification problems. With probabilistic classification being a subset of regression problems, this generalization opens up many new possibilities for the use of semi-supervised GANs as well as presenting an avenue for a deeper understanding of how they function. We first demonstrate the capabilities of semi-supervised regression GANs on a toy dataset which allows for a detailed understanding of how they operate in various circumstances. This toy dataset is used to provide a theoretical basis of the semi-supervised regression GAN. We then apply the semi-supervised regression GANs to the real-world application of age estimation from single images. We perform extensive tests of what accuracies can be achieved with significantly reduced annotated data. Through the combination of the theoretical example and real-world scenario, we demonstrate how semi-supervised GANs can be generalized to regression problems.

Assessing the fairness of a decision making system with respect to a protected class, such as gender or race, is challenging when class membership labels are unavailable. Probabilistic models for predicting the protected class based on observable proxies, such as surname and geolocation for race, are sometimes used to impute these missing labels for compliance assessments. Empirically, these methods are observed to exaggerate disparities, but the reason why is unknown. In this paper, we decompose the biases in estimating outcome disparity via threshold-based imputation into multiple interpretable bias sources, allowing us to explain when over- or underestimation occurs. We also propose an alternative weighted estimator that uses soft classification, and show that its bias arises simply from the conditional covariance of the outcome with the true class membership. Finally, we illustrate our results with numerical simulations and a public dataset of mortgage applications, using geolocation as a proxy for race. We confirm that the bias of threshold-based imputation is generally upward, but its magnitude varies strongly with the threshold chosen. Our new weighted estimator tends to have a negative bias that is much simpler to analyze and reason about.

Most of the existing work on automatic facial expression analysis focuses on discrete emotion recognition, or facial action unit detection. However, facial expressions do not always fall neatly into pre-defined semantic categories. Also, the similarity between expressions measured in the action unit space need not correspond to how humans perceive expression similarity. Different from previous work, our goal is to describe facial expressions in a continuous fashion using a compact embedding space that mimics human visual preferences. To achieve this goal, we collect a large-scale faces-in-the-wild dataset with human annotations in the form: Expressions A and B are visually more similar when compared to expression C, and use this dataset to train a neural network that produces a compact (16-dimensional) expression embedding. We experimentally demonstrate that the learned embedding can be successfully used for various applications such as expression retrieval, photo album summarization, and emotion recognition. We also show that the embedding learned using the proposed dataset performs better than several other embeddings learned using existing emotion or action unit datasets.

Every AI system is deployed by a human organization. In high risk applications, the combined human plus AI system must function as a high-reliability organization in order to avoid catastrophic errors. This short note reviews the properties of high-reliability organizations and draws implications for the development of AI technology and the safe application of that technology.