Artificial intelligence systems could outperform humans in all tasks within the next 45 years, according to a new study which also suggests that all human jobs will be automated in the next 120 years. According to a survey of over 350 artificial intelligence (AI) researchers, machines are predicted to be better than us at translating languages by 2024, writing high-school essays by 2026, driving a truck by 2027, working in retail by 2031, writing a bestselling book by 2049 and surgery by 2053. However, there is only a five per cent chance that computers will bring about outcomes that may lead to human extinction, researchers said. The survey, by the University of Oxford in the UK and Yale University in the US, was conducted among 352 researchers who had presented their research at the Conference on Neural Information Processing Systems or the International Conference on Machine Learning - the two major conferences in the field of AI. "There is accumulating evidence that machines can overpower human intelligence in complex, though specific tasks," Eleni Vasilaki at the University of Sheffield in the UK, told the'New Scientist'. However, there is little evidence that AI with human-like versatility will appear any time soon, Vasilaki said. The survey results showed that researchers in Asia typically gave shorter time frames than those in North America - predicting that AI would outperform humans on all tasks within 30 years, compared with 74 years.

With all the hype (and confusion) from vendors around AI, Gartner is taking the "bull" by the horns and launching a new collection of cool vendor reports just focused on AI. We have identified the most innovative emerging vendors with "cool" AI offerings that are having a real impact across different business functions, industry verticals, and geographies. This has meant sifting through the hundreds and hundreds of startups to find the ones that have a genuine AI offering (not simply slapping AI on their websites) and are really unique. Gartner defines AI through three key traits -- the ability to learn, predict and SURPRISE. Last week, my colleagues and I published our first "Cool Vendors in Core Technologies" research report highlighting five vendors that do indeed SURPRISE!

Uber said its head of finance is leaving, and the privately held ride-hailing company also said that its first-quarter loss narrowed substantially from the prior quarter, putting it on a path toward profitability. Head of finance Gautam Gupta is leaving in July to join another startup in San Francisco, the company said, making Gupta the latest high-profile executive to leave Uber. Uber, which has been rocked by several high-level executive departures in the past few months as it grapples with a series of controversies, has been looking for a chief operating officer to help change its now-notorious "bro" culture. Gupta's exit sets the stage for a second major executive search, now for a chief financial officer who has public company experience. About a dozen top executives have left Uber since February.

The goal of this article is to provide an overview of frameworks relevant to Machine Learning and Artificial Intelligence released by large corporations. We focus not just on pure Machine Learning and AI tools but also include some Big Data frameworks which provide value in making Machine Learning and AI available at scale. While these releases do have very strategic business reasons, there is no doubt that the trend of open-sourcing internal tools is adding value and making Machine Learning and AI more accessible. Over the past 2-3 years a large number of frameworks have been open-sourced. Companies may wish to establish standards, showcase their advanced level of research, attract talent or leverage the power of a community when open-sourcing tools. Whatever the reasons may be for open-sourcing tools, large organizations tend to have extensive resources which they use to build their internal tools. For businesses interested in exploring Data Science it only makes sense to evaluate whether any effort that has already gone into building these frameworks can be leveraged. We provide a summary of released tools, but not a comparison of the individual frameworks. Especially when it comes to Deep Learning, entire communities have formed around tools, and with that very dedicated fans and opponents. While we avoid such discussions, we provide our own observations and conclude the article with some generic guidelines for evaluating frameworks for business use.

The duopoly has gobbled up artificial intelligence startups, and it's hungry for more. Research released earlier this month by CB Insights found it acquired 34 AI startups in the first quarter of 2017, which more than doubled the total acquired during the same period last year and easily surpassed the previous high of 28, which came in the third quarter of last year. Last week, Axios reported that Google has launched an AI-specific venture fund, one that's overseen by engineers rather than venture capitalists, to hand out investments up to $10 million. The big picture • Google's parent, Alphabet, has acquired 11 AI startups over the past five years, the most of any public company. Apple is next, with seven, followed by Facebook and Intel, with five apiece.

You must understand algorithms to get good at machine learning. The problem is that they are only ever explained using Math. In this mega Ebook written in the friendly Machine Learning Mastery style that you're used to, finally cut through the math and learn exactly how machine learning algorithms work. Using clear explanations, simple pure Python code (no libraries!) and step-by-step tutorials you will discover how to load and prepare data, evaluate model skill, and implement a suite of linear, nonlinear and ensemble machine learning algorithms from scratch. I live in Australia with my wife and son and love to write and code.

From tingling in the limbs to increased heart rate and blood pressure - anxiety has been said to do a range of unusual things to the body. Now, a new study has found that this nervous disorder can also hinder your ability to interpret other people's emotions. Researchers have discovered that those in a heightened state of anxiety were unable to determine whether a person was happy or angry - and many people reported seeing the latter regardless of the facial expression. A new study has found that this nervous disorder can also hinder your ability to interpret facial expressions. Following the first two portions of the study, researcher had discovered that when individuals inhaled the carbon-dioxide rich air, or had an anxiety attack, they were eight percent worse at correctly identifying facial expressions.

In less than 50 years, artificial intelligence will be able to beat humans at all of their own tasks, according to a new study. And, the first hints of this shift will become apparent much sooner. Within the next ten years alone, the researchers found AI will outperform humans in language translation, truck driving, and even writing high-school essays – and, they say machines could be writing bestselling books by 2049. In less than 50 years, artificial intelligence will be able to beat humans at all of their own tasks, according to a new study. And, the first hints of this shift will become apparent much sooner.

Solving symmetric positive definite linear problems is a fundamental computational task in machine learning. The exact solution, famously, is cubicly expensive in the size of the matrix. To alleviate this problem, several linear-time approximations, such as spectral and inducing-point methods, have been suggested and are now in wide use. These are low-rank approximations that choose the low-rank space a priori and do not refine it over time. While this allows linear cost in the data-set size, it also causes a finite, uncorrected approximation error. Authors from numerical linear algebra have explored ways to iteratively refine such low-rank approximations, at a cost of a small number of matrix-vector multiplications. This idea is particularly interesting in the many situations in machine learning where one has to solve a sequence of related symmetric positive definite linear problems. From the machine learning perspective, such deflation methods can be interpreted as transfer learning of a low-rank approximation across a time-series of numerical tasks. We study the use of such methods for our field. Our empirical results show that, on regression and classification problems of intermediate size, this approach can interpolate between low computational cost and numerical precision.

The loss surface of deep neural networks has recently attracted interest in the optimization and machine learning communities as a prime example of high-dimensional non-convex problem. Some insights were recently gained using spin glass models and mean-field approximations, but at the expense of strongly simplifying the nonlinear nature of the model. In this work, we do not make any such assumption and study conditions on the data distribution and model architecture that prevent the existence of bad local minima. Our theoretical work quantifies and formalizes two important \emph{folklore} facts: (i) the landscape of deep linear networks has a radically different topology from that of deep half-rectified ones, and (ii) that the energy landscape in the non-linear case is fundamentally controlled by the interplay between the smoothness of the data distribution and model over-parametrization. Our main theoretical contribution is to prove that half-rectified single layer networks are asymptotically connected, and we provide explicit bounds that reveal the aforementioned interplay. The conditioning of gradient descent is the next challenge we address. We study this question through the geometry of the level sets, and we introduce an algorithm to efficiently estimate the regularity of such sets on large-scale networks. Our empirical results show that these level sets remain connected throughout all the learning phase, suggesting a near convex behavior, but they become exponentially more curvy as the energy level decays, in accordance to what is observed in practice with very low curvature attractors.