Amulya Bhuyan, 37 years old, lives in Dhowachala, in the northeastern state of Assam, and has few ways to buy new things. It takes hours to get to the nearest small town from the village of 1,000 people. Mr. Bhuyan, a teacher, made his first purchase on Amazon in 2016. After a recent delivery of a pair of jeans, he showed off other acquisitions: the shoes, socks, pants and shirt he was wearing; in his house, the curtains, glasses, flowery decals decorating the wall, a peacock clock and a painting of seven white horses running in the moonlight. "Before I didn't even know where to buy these things, and now they arrive on my doorstep," he said.

China's ever-growing reliance on facial recognition is spreading to public housing. Beijing is ramping up the use of face-detecting smart locks in public housing projects to bolster security for tenants (such as denying access to strangers) and crack down on abuses like illegal sublets. It even asks management to check on senior residents if they haven't entered or left their homes after a certain period of time. There are 47 projects using the technology as of the end of 2018, but the city now aims to have it in every project (serving about 120,000 tenants) by the end of June 2019. As with other uses of facial recognition in the country, this convenience comes at a steep cost to privacy. The building owners know exactly when tenants enter or leave, and the restrictions on strangers could make it difficult to invite guests without the building's knowledge.

John Marinos was skeptical when he received an email at his United Nations office in Bangkok that a team from SAP was developing an AI-based reporting tool to help this organization better manage humanitarian aid. One year later he's convinced. The aptly named 4W-Wizard has turned thousands of lines of non-standard, confusing and misspelled information from numerous agencies on the ground into clean data that the United Nations Office for the Coordination of Humanitarian Affairs (UN OCHA) can use for streamlined disaster relief management. AI-based tool helps support UN's vision of a seamless system across the humanitarian universe of support.Getty Images "Since 90 percent of our challenges involve people-oriented problems, I never believed technology would help us to sort this out," said Marinos, information management officer at UN OCHA. "We've struggled with manual-based reporting that took hours to clean up before it was usable. The 4W-Wizard helps provide fast visibility into which organizations are providing what kind of help where and when. In less than an hour, we can immediately see where the gaps are and what's needed next."

W. Edward Deming taught that quality is achieved by measuring as much as possible and reducing variations, and reducing variation is achieved by improving the system, not just pieces. Japan widely adopted Deming's philosophies in the 1950s and became the 2nd biggest economy in the world. Quality improvement didn't decrease jobs in Japan, it increased jobs. AI now has the ability to expand and codify Deming's philosophies – to take them to the next level. AI can improve and standardize decision making based on logic, rather than the fear of missing objectives, bonuses or losing one's job.

Mr Sandeep Parikh, Intelligent Automation Partner at EY talks about hoe AI and ML will be utilized in India, some advancement and use cases of the same and investments made into the technology. Over the past several months service providers (baring a few) have rushed to market with a wide range of Artificial Intelligence and Machine Learning based home-grown solution. However, there is (by and large) a lack of awareness in the market place of how to apply these technologies in the real world to generate a relevant and lasting impact. So I believe, 2019 will see providers, build stronger capability, focus on relevant real industry scenarios and hence success stories. There is still some time to go in terms of creating concrete and end-to-end true differentiation using cognitive solutions – while we have seen some interesting pieces in operations, FS front offices, finance, HR and marketing where ML has potentially created small but real impact e.g.

Some surveys estimate that one in ten people, or about 40 million Americans and 700 million children and adults worldwide, have dyslexia, and according to the Dyslexia Center of Utah, 70 to 80 percent of people with poor reading skills are likely dyslexic. It's hardly a death sentence, but if left untreated, dyslexia can severely impede tasks like organization, planning and prioritizing, and keeping time. Artificial intelligence (AI) might someday lend a helping hand, hopefully. Researchers at the Israel Institute of Technology's Laboratory of Clinical Neurophysiology and the University of Haifa's Department of Computer Science claim to have developed an AI model that can automatically, without human intervention and with state-of-the-art precision, identify dyslexic readers. They describe their work in a preprint paper ("Features and Machine Learning for Correlating and Classifying between Brain Areas and Dyslexia") published on Arxiv.org this week.

Rapid tech transformation in the business landscape was the highlight of 2018. Business processes across industries were overhauled almost overnight on the back of artificial intelligence, machine learning, blockchain, robotics and data analytics. This transformation changed the nature of jobs, making them more tech-centric than ever before. As companies begin hiring technologically-adept individuals, here is a look at the jobs that will be most popular in 2019. In India, according to a Nasscom report, the data analytics sector will rise to $16 billion by 2025.

Asynchronous momentum stochastic gradient descent algorithms (Async-MSGD) have been widely used in distributed machine learning, e.g., training large collaborative filtering systems and deep neural networks. Due to current technical limit, however, establishing convergence properties of Async-MSGD for these highly complicated nonoconvex problems is generally infeasible. Therefore, we propose to analyze the algorithm through a simpler but nontrivial nonconvex problems --- streaming PCA. This allows us to make progress toward understanding Aync-MSGD and gaining new insights for more general problems. Specifically, by exploiting the diffusion approximation of stochastic optimization, we establish the asymptotic rate of convergence of Async-MSGD for streaming PCA. Our results indicate a fundamental tradeoff between asynchrony and momentum: To ensure convergence and acceleration through asynchrony, we have to reduce the momentum (compared with Sync-MSGD). To the best of our knowledge, this is the first theoretical attempt on understanding Async-MSGD for distributed nonconvex stochastic optimization. Numerical experiments on both streaming PCA and training deep neural networks are provided to support our findings for Async-MSGD.

The backpropagation of error algorithm (BP) is impossible to implement in a real brain. The recent success of deep networks in machine learning and AI, however, has inspired proposals for understanding how the brain might learn across multiple layers, and hence how it might approximate BP. As of yet, none of these proposals have been rigorously evaluated on tasks where BP-guided deep learning has proved critical, or in architectures more structured than simple fully-connected networks. Here we present results on scaling up biologically motivated models of deep learning on datasets which need deep networks with appropriate architectures to achieve good performance. We present results on the MNIST, CIFAR-10, and ImageNet datasets and explore variants of target-propagation (TP) and feedback alignment (FA) algorithms, and explore performance in both fully- and locally-connected architectures. We also introduce weight-transport-free variants of difference target propagation (DTP) modified to remove backpropagation from the penultimate layer. Many of these algorithms perform well for MNIST, but for CIFAR and ImageNet we find that TP and FA variants perform significantly worse than BP, especially for networks composed of locally connected units, opening questions about whether new architectures and algorithms are required to scale these approaches. Our results and implementation details help establish baselines for biologically motivated deep learning schemes going forward.

Asynchronous momentum stochastic gradient descent algorithms (Async-MSGD) have been widely used in distributed machine learning, e.g., training large collaborative filtering systems and deep neural networks. Due to current technical limit, however, establishing convergence properties of Async-MSGD for these highly complicated nonoconvex problems is generally infeasible. Therefore, we propose to analyze the algorithm through a simpler but nontrivial nonconvex problems --- streaming PCA. This allows us to make progress toward understanding Aync-MSGD and gaining new insights for more general problems. Specifically, by exploiting the diffusion approximation of stochastic optimization, we establish the asymptotic rate of convergence of Async-MSGD for streaming PCA. Our results indicate a fundamental tradeoff between asynchrony and momentum: To ensure convergence and acceleration through asynchrony, we have to reduce the momentum (compared with Sync-MSGD). To the best of our knowledge, this is the first theoretical attempt on understanding Async-MSGD for distributed nonconvex stochastic optimization. Numerical experiments on both streaming PCA and training deep neural networks are provided to support our findings for Async-MSGD.