Artificial intelligence (AI) on its own isn't enough to compete -- companies need industry-specific solutions to business problems. So said Martin Schroeter, IBM Corp. (NYSE: IBM)'s company senior vice president and chief financial officer, on the company's quarterly earnings call Thursday afternoon. Cognitive computing technology (IBM's term for AI) is just "table stakes," said Schroeter, claiming that his company is going the extra mile. IBM is building datasets for Watson to serve specific industries, including healthcare and finance. "You need more than public data or algorithms to solve real-world problems," Schroeter said.

An ongoing challenge in neuromorphic computing is to devise general and computationally efficient models of inference and learning which are compatible with the spatial and temporal constraints of the brain. One increasingly popular and successful approach is to take inspiration from inference and learning algorithms used in deep neural networks. However, the workhorse of deep learning, the gradient descent Back Propagation (BP) rule, often relies on the immediate availability of network-wide information stored with high-precision memory, and precise operations that are difficult to realize in neuromorphic hardware. Remarkably, recent work showed that exact backpropagated weights are not essential for learning deep representations. Random BP replaces feedback weights with random ones and encourages the network to adjust its feed-forward weights to learn pseudo-inverses of the (random) feedback weights. Building on these results, we demonstrate an event-driven random BP (eRBP) rule that uses an error-modulated synaptic plasticity for learning deep representations in neuromorphic computing hardware. The rule requires only one addition and two comparisons for each synaptic weight using a two-compartment leaky Integrate & Fire (I&F) neuron, making it very suitable for implementation in digital or mixed-signal neuromorphic hardware. Our results show that using eRBP, deep representations are rapidly learned, achieving nearly identical classification accuracies compared to artificial neural network simulations on GPUs, while being robust to neural and synaptic state quantizations during learning.

In many applications of black-box optimization, one can evaluate multiple points simultaneously, e.g. when evaluating the performances of several different neural network architectures in a parallel computing environment. In this paper, we develop a novel batch Bayesian optimization algorithm --- the parallel knowledge gradient method. By construction, this method provides the one-step Bayes optimal batch of points to sample. We provide an efficient strategy for computing this Bayes-optimal batch of points, and we demonstrate that the parallel knowledge gradient method finds global optima significantly faster than previous batch Bayesian optimization algorithms on both synthetic test functions and when tuning hyperparameters of practical machine learning algorithms, especially when function evaluations are noisy.

In this paper, a genetic algorithm-based frequency-domain feature search (GAFDS) method is proposed for the electroencephalogram (EEG) analysis of epilepsy. In this method, frequency-domain features are first searched and then combined with nonlinear features. Subsequently, these features are selected and optimized to classify EEG signals. The extracted features are analyzed experimentally. The features extracted by GAFDS show remarkable independence, and they are superior to the nonlinear features in terms of the ratio of inter-class distance and intra-class distance. Moreover, the proposed feature search method can additionally search for features of instantaneous frequency in a signal after Hilbert transformation. The classification results achieved using these features are reasonable, thus, GAFDS exhibits good extensibility. Multiple classic classifiers (i.e., $k$-nearest neighbor, linear discriminant analysis, decision tree, AdaBoost, multilayer perceptron, and Na\"ive Bayes) achieve good results by using the features generated by GAFDS method and the optimized selection. Specifically, the accuracies for the two-classification and three-classification problems may reach up to 99% and 97%, respectively. Results of several cross-validation experiments illustrate that GAFDS is effective in feature extraction for EEG classification. Therefore, the proposed feature selection and optimization model can improve classification accuracy.

If you use your credit card to buy a latte in Vancouver and a couple of minutes later that card is making a purchase in Singapore, that's a red flag for fraud. But increasingly sophisticated fraud calls for more sophisticated measures to deal with it, and that is among the challenges behind RBC Research's announcement today that it is launching a new lab to explore the use of artificial intelligence and machine learning in the financial sector. Richard Sutton, a computer scientist and pioneer in artificial intelligence, has been named head academic advisor to RBC Research in machine learning. The new lab will work with the Alberta Machine Intelligence Institute at the University of Alberta, where Sutton is a professor. Foteini Agrafioti, head of RBC Research, which was launched last fall in Toronto, said the announcement will help her organization to play a major role in advancing AI research in the future of banking. Agrafioti said that as the complexity of fraud evolves over time, it becomes increasingly difficult to detect it.

Scientists have pulled inspiration from snakes to help robots detect when humans are near. The heat-sensing film has the ability to detect tiny temperature changes, and was inspired in the sensors a viper uses its skin to find prey. Researchers foresee this technology being used by search and rescue robots in order to locate humans in disaster zones. Scientists have pulled inspiration from snakes to help robots detect when humans are near. The heat-sensing film is a flexible, transparent coating made of pectin.

Elon Musk was right when he said that Artificial Intelligence (AI) systems and automation will eventually take over most human jobs and, in what can be seen as a first case in India where the country's software giant Infosys has "released" 8,000-9,000 employees in the past one year because of automation of lower-end jobs, which is confirmed by the company's human resources head Krishnamurthy Shankar in a statement to ET. Notably, these released employees are now working on more advanced projects, so its not that automation or AI has stole the human jobs but it actually transforming at least in this case of Infosys. Machine learning and Artificial Intelligence previously required expensive computing machines, but costs have now come down dramatically. As part of its Automation, Artificial Intelligence and machine learning adaptation drive, the company has been releasing about 2,000 people every quarter and also training them in special courses that will help them in their new assignments. In a research report, released by Infosys itself said that the majority -- 85 percent -- plan to train employees about the benefits and use of AI, and 80 percent of companies replacing roles with AI technologies will retrain or redeploy displaced employees. The report also predicts that organizations that have already deployed or have plans to deploy AI technologies expect to see a 39 percent average increase in revenue by 2020, alongside a 37 percent reduction in costs.

Machine learning has come a long way since Hollywood painted it as shiny robots fueled by artificial intelligence. In the Hollywood version, robots usually end up replacing humans. But today, we're actually using machine learning to supplement many of the things that humans do best. They feel foreign, scientific, and hard to understand. And for many professionals, the phrase still sounds like highly technical jargon.

"Accelerating artificial intelligence (AI) capabilities will enable automation of some tasks that have long required human labor," notes the White House report in its opening paragraph. "These transformations will open up new opportunities for individuals, the economy, and society, but they have the potential to disrupt the current livelihoods of millions of Americans. Whether AI leads to unemployment and increases in inequality over the long-run depends not only on the technology itself but also on the institutions and policies that are in place." How strongly will AI disrupt the US workforce? Is this time different from past technological disruptions?

How can I become a data scientist? A quick search yields a plethora of possible resources that could help -- MOOCs, blogs, Quora answers to this exact question, books, Master's programs, bootcamps, self-directed curricula, articles, forums and podcasts. Their quality is highly variable; some are excellent resources and programs, some are click-bait laundry lists. Since this is a relatively new role and there's no universal agreement on what a data scientist does, it's difficult for a beginner to know where to start, and it's easy to get overwhelmed. Many of these resources follow a common pattern: 1) Here are the skills you need and 2) Here is where you learn each of these.