On March 7, hackers engaged in cryptocurrency mining, attacked one of the largest torrent-applications for downloading movies (the second in the world after BitTorrent), well known, including in Russia as Mediaget.com. Criminals managed to hack the server through which the torrent update is carried out and the "secret miner" was installed on the computers of Mediaget customers, which began to actively absorb the resources of users' electronic devices. According to Microsoft, who managed to stop the invasion in time, if not for the qualified actions of their employees, the number of victims around the world would have been estimated in the millions. The calculation of criminals was simple and was based on the desire to obtain basic physical resources to expand the scale of their own business. In this regard, it should be noted that the production of any electronic currency requires two main components - computer hardware with powerful graphics cards and, as a result, a large amount of electricity.

Artificial intelligence and robotics are disrupting every aspect of work and redefining productivity. The old ways of not just working, but also assessing capabilities, hiring and compensation, are undergoing a massive change. In a conversation with Knowledge@Wharton, Srikanth Karra, chief human resource officer at Indian IT services firm Mphasis, discusses what this means for individuals, organizations and countries. Karra said managerial jobs and tasks that are repetitive in nature will be displaced and the ability to learn new skills will be critical for individuals who want to stay relevant. Companies will need to devise new ways of training and assessing the skills of employees while countries must develop a learning ecosystem. "Work will be more contractual in nature and deep technical skills, creativity and learnability will be at a premium," he noted.

Language and conversation are some of the most fundamental technologies that still exist today. These technologies have been evolving over thousands of years. Yet when it comes to designing a conversation we struggle to understand just how complicated conversation really is. How do we understand language? How do we form words? How do we hold conversations?

The coverage and hype around artificial intelligence (AI) is reaching fever pitch. Key questions explore how it might impact people's lives and employment over the next five to ten years. These lead to questions over how to fund tax revenue shortfalls and higher unemployment costs. In practice, it's too early to know how fast AI will advance, how far it will spread into society, whether it will reach a state of superintelligence where it outsmarts humans at everything, or the net employment impact. This article explores the key economic questions that arise around the potential impact of AI on jobs in society, with a focus on robot taxes as a way of funding higher unemployment benefit costs or Guaranteed / Universal Basic Income (UBI) and Universal Basic Services (UBS) schemes.

Plenty of cars will help you park, but the biggest challenge is frequently finding a spot in the first place -- it's no fun to circle the parking lot for 10 minutes. Fully autonomous cars can ultimately take care of this, but Jaguar Land Rover is demonstrating a feature that would help in the meantime. It recently expanded its public semi-autonomous testing in the UK to include a "self-driving valet" where vehicles both find open spaces and park themselves. The company pitches it as eliminating some of the drudgery of driving, letting you take the wheel when you'd genuinely enjoy it. The automaker has also been testing other connected car features, including a collaborative parking feature where vehicles share info about free spaces as you approach a parking lot.

Self-driving car trials are to continue in the UK despite mounting concerns over safety after an Uber autonomous vehicle struck and killed a US pedestrian in Arizona this week. The country's biggest carmaker, Jaguar Land Rover, pushed ahead with trials of its autonomous vehicles in the Midlands yesterday despite warnings that the public are being treated like'human guinea pigs' during driverless car tests. The trial, launched less than 48 hours after the fatal accident on Sunday, is believed to be the first time a self-driving car has been used on open, public roads. The firm is expected to demonstrate more of the cars' features, including an emergency braking system, on urban streets in further tests this week. Britain's biggest carmaker, Jaguar Land Rover, pushed ahead with trials of its autonomous vehicles (file photo) in the Midlands yesterday despite warnings that the public are being treated like'human guinea pigs' during driverless car tests A self-driving Range Rover Sport drove itself through the centre of Milton Keyes on Tuesday before parking and driving off again, as part of a government-backed trial.

"Reducing the everyday stresses of driving - like squeezing into a tight parking place - means that we can all focus on the more enjoyable aspects of our cars," said Joerg Schlinkheider, Jaguar Land Rover Chief Engineer, Automated Driving.

Interventional clinical cancer trials are generally too restrictive and cancer patients are often excluded from them on the basis of comorbidity, past or concomitant treatments and the fact that they are over a certain age. The efficacy and safety of new treatments for patients with these characteristics are not, therefore, defined. In this work, we build a model with which to automatically predict whether short clinical statements were considered inclusion or exclusion criteria. We used clinical trials protocols on cancer that have been available in public registries for the last 18 years to train word embeddings, and constructed a dataset of 6M short free-texts labeled as eligible or not eligible. We then trained and validated a text classifier, using deep neural networks with pre-trained word-embedding as its inputs, to predict whether or not short free-text statements describing clinical information were considered eligible. The best model achieved an F-measure of 0.92 and an almost perfect agreement when employing a validation set of 800K labeled statements. The trained model was also tested on an independent set of clinical statements mimicking those used in routine clinical practice, yielding a consistent performance. We additionally analyzed the semantic reasoning of the word embedding representations obtained, and were able to identify equivalent treatments for a type of tumor in an analogy with the drugs used to treat other tumors. The present work shows that representation learning using neural networks can be successfully leveraged to extract the medical knowledge available on clinical trial protocols and potentially assist practitioners when prescribing treatments.

Measuring uncertainty is a promising technique for detecting adversarial examples, crafted inputs on which the model predicts an incorrect class with high confidence. But many measures of uncertainty exist, including predictive en- tropy and mutual information, each capturing different types of uncertainty. We study these measures, and shed light on why mutual information seems to be effective at the task of adversarial example detection. We highlight failure modes for MC dropout, a widely used approach for estimating uncertainty in deep models. This leads to an improved understanding of the drawbacks of current methods, and a proposal to improve the quality of uncertainty estimates using probabilistic model ensembles. We give illustrative experiments using MNIST to demonstrate the intuition underlying the different measures of uncertainty, as well as experiments on a real world Kaggle dogs vs cats classification dataset.

The stochastic gradient descent (SGD) optimization algorithm plays a central role in machine learning and, in particular, deep learning applications such as image analysis and speech recognition (cf., e.g., [12, 13, 16, 23]). It is therefore important to analyze and quantify the convergence speed of the SGD method. There is a vast amount of scientific literature investigating and providing upper bounds for the SGD method and modifications of it (cf., e.g., [3, 4, 5, 6, 7, 8, 9, 10, 11, 18, 20, 21, 24] and cf., e.g., [14] for a more comprehensive review of the literature). Much less attention has been paid to proving lower error bounds for the SGD method, that is, to quantifying the best possible speed of convergence which the SGD method can achieve (cf., e.g., [2, 17, 19, 22, 25]). It is the key contribution of this paper to make a step in this direction.