The development and adoption of advanced technologies including smart automation and artificial intelligence has the potential not only to raise productivity and GDP growth but also to improve well-being more broadly, including through healthier life and longevity and more leisure. Alongside such benefits, these technologies also have the potential to reduce disruption and the potentially destabilizing effects on society arising from their adoption. Tech for Good: Smoothing disruption, improving well-being (PDF–1MB) examines the factors that can help society achieve such benefits and makes a first attempt to calculate the impact of technology adoption on welfare growth beyond GDP. Our modeling suggests that good outcomes for the economy overall and for individual well-being come about when technology adoption is focused on innovation-led growth rather than purely on labor reduction and cost savings through automation. This needs to be accompanied by proactive transition management that increases labor market fluidity and equips workers with new skills. Technology for centuries has both excited the human imagination and prompted fears about its effects. Today's technology cycle is no different, provoking a broad spectrum of hopes and fears. Opinion surveys suggest people tend to have a nuanced view of technology but nonetheless worry about the risks: while generally positive about longer-term benefits, especially for health, many are also concerned about the negative impact on their lives, in particular in the areas of job security, material living standards, safety, and trust.

Bengaluru: Disruptive technologies and Artificial Intelligence (AI) are making their way into classrooms as humanoid robots to teach students and interact with them as teachers do, at a school in Bengaluru. "Our robots impart lessons daily in five subjects to about 300 students in Classes 7-9 in four sections by turns. They also interact with them and respond to questions in the subjects," Indus International School's chief design officer, Vignesh Rao, told IANS here. Though the 5 foot 7 inch robots, dressed in formal female attire, do not replace real teachers, they complement them in teaching lessons in the subjects and reply to FAQs (frequently asked questions) from students. "We have programmed the interactive robots to answer questions students frequently ask on the subjects and related to them. With AI in play, the robots are able to respond to questions and doubts of our wards after a lesson is taught," said Rao.

Although corporate leaders have talked about skills gaps for years, the spread of automation and artificial intelligence is prompting some of the biggest companies -- including Amazon, JPMorgan Chase, SAP, Walmart, and AT&T, to name just a few -- to take action, not with small pilots but with comprehensive plans to retrain large segments of their workforces. These programs signal that the "future of work" is no longer an event on the distant horizon. Our latest research finds that the occupational mix of the economy is already shifting in ways that will accelerate over the next decade. Although we estimate that only 5% of all occupations can be fully automated, the activities in nearly all jobs will evolve. As intelligent machines take over many physical, repetitive, or basic cognitive tasks, the work that remains will involve both more technical and digital skills and more personal interaction, creativity, and judgment.

Artificial intelligence has thrashed humans at chess. Now the bots are gunning for mahjong. An AI-powered program developed by Microsoft Corp. has surpassed the average level of the top players in a recent competition in Japan, Harry Shum, executive vice-president of the company's artificial intelligence and research group, said in Shanghai on Thursday. "To those friends who usually lose money in mahjong, this is good news to you," Shum said to laughter at the World AI Conference. "The bot player developed by Microsoft can deal with high uncertainty, presenting instincts akin to human, projection and deduction capabilities as well as a sense of overall consciousness."

Emily He, senior vice president, Oracle HCM Cloud shares her thoughts on reskilling employees for their AI promotion. It's easy to be impressed by artificial intelligence's significant role in simplifying and streamlining work. It's the same reason many are afraid of AI: Will it lead to fewer jobs and human interaction? These fears are pervasive, but the reality is that when harnessed appropriately, AI does the exact opposite and (as counterintuitive as it may sound) can actually make work more human. Yes, artificially intelligent solutions can automate mundane tasks, save time and resources, and cut costs.

About a month ago in Pittsburgh, we brought together a group of educators, students, and technology users to ponder some big unwieldy questions about the intersection of education, artificial intelligence, and youth media. But instead of having attendees passively ingest ideas from a speaker on stage, the event started by giving every guest a chance to pause, think, and respond to a series of questions. How often do we draw the line between what we can do with tech to what we should do with it? I see a lot of focus on the "risk" of online spaces. When profit is the motivation of AI systems, there is always an overlooking of what people actually need and/or want.

Technologies such as artificial intelligence (AI) and virtual reality (VR) are rapidly expanding opportunities for teaching and learning, and they are giving college administrators new and different ways to track student outcomes. To learn more about the impact of these technologies, we attended a handful of panels on the topic led by higher education and technology leaders at Educause's annual conference in Denver this week. From teaching with VR to tracking student success with AI, we explore how colleges and universities are using new technologies to conduct research, teach students and create smarter campuses. Virtual and augmented reality tools can provide students with experiences that would be otherwise too expensive or even impossible to replicate in the real world, from exploring the inside of a cell to traversing faraway planets, said D. Christopher Brooks, director of research at the Educause Center for Analysis and Research. At Hamilton College, for example, these tools are changing the way the 1,850-student liberal arts institution teaches human anatomy.

Knowledge distillation (KD) is a very popular method for model size reduction. Recently, the technique is exploited for quantized deep neural networks (QDNNs) training as a way to restore the performance sacrificed by word-length reduction. KD, however, employs additional hyper-parameters, such as temperature, coefficient, and the size of teacher network for QDNN training. We analyze the effect of these hyper-parameters for QDNN optimization with KD. We find that these hyper-parameters are inter-related, and also introduce a simple and effective technique that reduces \textit{coefficient} during training. With KD employing the proposed hyper-parameters, we achieve the test accuracy of 92.7% and 67.0% on Resnet20 with 2-bit ternary weights for CIFAR-10 and CIFAR-100 data sets, respectively.

Prior work poses a diversity paradox for science. Diversity breeds scientific innovation, and yet, diverse individuals have less successful scientific careers. But if diversity is good for innovation, why is science not rewarding diversity? We answer this question by utilizing a near-population of ~1.03 million US doctoral recipients from 1980-2015 and their careers into publishing and faculty roles. The article uses text analysis and machine learning techniques to answer a series of questions: How can we detect scientific innovation? Does diversity breed innovation? And are the innovations of diverse individuals adopted and rewarded? Our analyses show that underrepresented groups produce higher rates of scientific novelty. However, their novel contributions are discounted: e.g., innovations by gender minorities are taken up by other scholars at lower rates than innovations by gender majorities, and innovations by gender and racial minorities result in fewer academic positions. This suggests an unfair system in which diverse individuals innovate, but their innovations are disproportionately ignored and fail to convert into career success at the same rate as majority groups. In sum, there may be an unwarranted reproduction of stratification in academic careers that discounts diversity's role in innovation and partly explains the underrepresentation of some groups in academia.

Deep learning algorithms often require solving a highly non-linear and nonconvex unconstrained optimization problem. Methods for solving optimization problems in large-scale machine learning, such as deep learning and deep reinforcement learning (RL), are generally restricted to the class of first-order algorithms, like stochastic gradient descent (SGD). While SGD iterates are inexpensive to compute, they have slow theoretical convergence rates. Furthermore, they require exhaustive trial-and-error to fine-tune many learning parameters. Using second-order curvature information to find search directions can help with more robust convergence for non-convex optimization problems. However, computing Hessian matrices for large-scale problems is not computationally practical. Alternatively, quasi-Newton methods construct an approximate of the Hessian matrix to build a quadratic model of the objective function. Quasi-Newton methods, like SGD, require only first-order gradient information, but they can result in superlinear convergence, which makes them attractive alternatives to SGD. The limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) approach is one of the most popular quasi-Newton methods that construct positive definite Hessian approximations. In this chapter, we propose efficient optimization methods based on L-BFGS quasi-Newton methods using line search and trust-region strategies. Our methods bridge the disparity between first- and second-order methods by using gradient information to calculate low-rank updates to Hessian approximations. We provide formal convergence analysis of these methods as well as empirical results on deep learning applications, such as image classification tasks and deep reinforcement learning on a set of ATARI 2600 video games. Our results show a robust convergence with preferred generalization characteristics as well as fast training time.