Self-driving still seems to be a ways off from active public use on regular roads, but once it arrives, it could ramp very quickly, according to a new study by the Boston Consulting Group. The study found that by 2030, up to a quarter of driving miles in the U.S. could be handled by self-driving electric vehicles operating in shared service fleets in cities, due mostly to considerable cost savings for urban drivers. The big change BCG sees is a result of the rise in interest in autonomous technologies, paired with the increased electrification of vehicles. There's also more pressure on cities to come up with alternate transportation solutions that address increasing congestion. All of that added together could drive reduction in costs by up to 60 percent for drivers who opt into using shared self-driving services vs. owning and operating their own cars.

Machine learning is a part of everyday life for most Americans, from navigation apps to Amazon's omniscient purchase recommendations. But in healthcare the use of machine learning has so far been limited to niche science projects in large and academic health systems – those able to afford the highly skilled data scientists and dedicated teams required to turn their data into meaningful performance improvements. Health Catalyst is on a mission to change that by embedding the value of machine learning throughout healthcare. Last month, the company launched healthcare.ai to help make machine learning routine, pervasive and actionable for healthcare organizations of all sizes. The collaborative, open source repository of machine learning tools and expertise including topical blog content and weekly live hands-on machine learning educational broadcasts, makes it easy to deploy machine learning in any environment.

In recent years, we have become increasingly good at training deep neural networks to learn a very accurate mapping from inputs to outputs, whether they are images, sentences, label predictions, etc. from large amounts of labeled data. What our models still frightfully lack is the ability to generalize to conditions that are different from the ones encountered during training. Every time you apply your model not to a carefully constructed dataset but to the real world. The real world is messy and contains an infinite number of novel scenarios, many of which your model has not encountered during training and for which it is in turn ill-prepared to make predictions. The ability to transfer knowledge to new conditions is generally known as transfer learning and is what we will discuss in the rest of this post. Over the course of this blog post, I will first contrast transfer learning with machine learning's most pervasive and successful paradigm, supervised learning. I will then outline reasons why transfer learning warrants our attention. Subsequently, I will give a more technical definition and detail different transfer learning scenarios.

As it has more than 12,500 patents, eight Nobel prizes, and a 140-year history of field-testing crazy ideas, it should surprise no one that AT&T would be an important player in artificial intelligence. "AT&T is a backbone of the internet," explains Nadia Morris, head of Innovation at the AT&T Connected Health Foundry. The company manages wireless, landline, and even private secure networks to power connectivity for both individuals and corporations. All these networks generate incredible volumes of data that is ripe for machine analysis. AT&T has built AI and machine learning systems for decades, using algorithms to automate operations such as common call center procedures and the analysis and correction of network outages.

Beijing: Alan Du, a venture capitalist and World Series of Poker veteran, was in his fifth day of matching wits against his stone-cold opponent-- and his losses were piling up. His rival was literally inhuman. That's because Du went up against "Lengpudashi" an updated version of the Libratus artificial intelligence program that achieved a major milestone by besting four of the world's best poker pros in January. Housed within a super-computing centre near Carnegie Mellon University in Pittsburgh, its name, intended to resemble its English moniker, fittingly translates into "cold poker master." Du and five team members played 36,000 hands against the machine over the course of five days.

Let's take a breath: Robots and artificial intelligence systems are nowhere near displacing the human workforce. Nevertheless, no less a voice than Bill Gates has asserted just the opposite and called for a counterintuitive, preemptive strike on these innovations. His proposed weapon of choice? Taxes on technology to compensate for losses that haven't happened. David Kenny (@davidwkenny) is IBM's senior vice president for Watson and the company's cloud platform.

Whenever the term "artificial intelligence" is mentioned, the first thought that comes to any layman's mind is that it is the technology which helps develop robots. Major Hollywood movies have given a superficial impression about the technological subject. But artificial intelligence is a much broader concept that the regular notion of it. If we talk about it in the literal sense, the term artificial intelligence stands for the simulation of human intelligence processed by machines, mainly computer systems. The major processes under this simulation are reasoning, learning, and self-correction. These processes are later elaborated with reference to the needs of technology.

The ritualized procreation in the novel--effectively, state-sanctioned rape--is extrapolated from the Bible. " 'Behold my maid Bilhah, go in unto her; and she shall bear upon my knees, that I may also have children by her,' " Atwood recited. "Obviously, they stuck the two together and out came the baby, and it was given to Rachel.

We introduce a novel approach for estimating Latent Dirichlet Allocation (LDA) parameters from collapsed Gibbs samples (CGS), by leveraging the full conditional distributions over the latent variable assignments to efficiently average over multiple samples, for little more computational cost than drawing a single additional collapsed Gibbs sample. Our approach can be understood as adapting the soft clustering methodology of Collapsed Variational Bayes (CVB0) to CGS parameter estimation, in order to get the best of both techniques. Our estimators can straightforwardly be applied to the output of any existing implementation of CGS, including modern accelerated variants. We perform extensive empirical comparisons of our estimators with those of standard collapsed inference algorithms on real-world data for both unsupervised LDA and Prior-LDA, a supervised variant of LDA for multi-label classification. Our results show a consistent advantage of our approach over traditional CGS under all experimental conditions, and over CVB0 inference in the majority of conditions. More broadly, our results highlight the importance of averaging over multiple samples in LDA parameter estimation, and the use of efficient computational techniques to do so.

We formulate the notion of minimax estimation under storage or communication constraints, and prove an extension to Pinsker's theorem for nonparametric estimation over Sobolev ellipsoids. Placing limits on the number of bits used to encode any estimator, we give tight lower and upper bounds on the excess risk due to quantization in terms of the number of bits, the signal size, and the noise level. This establishes the Pareto optimal tradeoff between storage and risk under quantization constraints for Sobolev spaces. Our results and proof techniques combine elements of rate distortion theory and minimax analysis. The proposed quantized estimation scheme, which shows achievability of the lower bounds, is adaptive in the usual statistical sense, achieving the optimal quantized minimax rate without knowledge of the smoothness parameter of the Sobolev space. It is also adaptive in a computational sense, as it constructs the code only after observing the data, to dynamically allocate more codewords to blocks where the estimated signal size is large. Simulations are included that illustrate the effect of quantization on statistical risk.