Climate action is the latest buzzword among industry circles since the many International Panel on Climate Change (IPCC) reports and the recent UN Climate Summit in New York City. Greta Thunberg grabbed the headlines, but industrialists are all wondering: How can we move swiftly and effectively to reduce carbon emissions? How can we use AI and other exponential technologies to do the job better, faster and cheaper? As a business strategist and urban planner, I advise companies to focus on cities since they consume 80% of energy and emit 70% of carbon, so we'll win or lose the carbon battle in the cities. Fortunately, cities can move faster than national governments and, as energy buyers, they can directly negotiate energy types and pricing, giving them enormous economic clout.

In Part 1 of this 3-part miniseries, we discussed the origins of artificial intelligence (AI), and we considered some low-hanging AI-enabled fruit in the form of speech recognition, voice control, and machine vision. In Part 2, we noted some of the positive applications of AI, like recognizing skin cancer, identifying the source of outbreaks of food poisoning, and the early detection of potential pandemics. In fact, there are so many feel-good possibilities for the future that they can make your head spin. In a moment, we'll ponder a few more of these before turning our attention to the dark side. Another topic we considered in Part 2 was the combination of mediated reality (MR) and AI, where mediated reality encompasses both augmented reality (AR) and deletive reality (DR). In the case of AR, information is added to the reality we are experiencing.

Researchers from The University of Western Australia and the University of California have developed a new machine-learning tool that will improve the management, restoration and irrigation of rangeland areas used for grazing. Associate professor Sally Thompson from the UWA School of Engineering and UWA Institute of Agriculture says the tool was suited to environments where the amount of rainfall exceeded the absorption capacity of the soil, resulting in the excess water flowing over the land. The new machine learning tool models surface water flows in dry environments with patchy vegetation cover. The findings have significant implications for agricultural and natural systems in Australia and worldwide. "The research could help environmental designers limit soil erosion in rangeland environments and agricultural systems, minimising degradation risks in drylands," says Sally. "It also has applications in urban settings, where waterproof surfaces, like pavement, generate runoff and flood risks."

Australia is home to more than 10,000 beaches, ranging from a few dozen metres to hundreds of kilometres long. But increasingly, these beaches are slowly disappearing before our eyes. "Beaches across Australia are eroding, simply because waves come in pull sand away -- and big storm surges pull more sand away," IBM Systems Data Scientist Dr Adam Makarucha told the Gartner Application Architecture, Development, and Integration Summit in Sydney. While the likes of Gold Coast Council have invested AU$14 million into rehabilitation projects -- such as one for a 12km stretch of beach, equating to more than a million dollars per kilometre -- Makarucha said prevention is more viable than rehabilitation. Makarucha said the best way to prevent beach erosion is to look to a natural defence, such as seagrass.

The convergence of AI and HPC has created a fertile venue that is ripe for imaginative researchers -- versed in AI technology -- to make a big impact in a variety of scientific fields. From new hardware to new computational approaches, the true impact of deep- and machine learning on HPC is, in a word, "everywhere". Just as technology changes in the personal computer market brought about a revolution in the design and implementation of the systems and algorithms used in high performance computing (HPC), so are recent technology changes in machine learning bringing about an AI revolution in the HPC community. Expect new HPC analytic techniques including the use of GANs (Generative Adversarial Networks) in physics-based modeling and simulation, as well as reduced precision math libraries such as NLAFET and HiCMA to revolutionise many fields of research. Other benefits of the convergence of AI and HPC include the physical instantiation of data flow architectures in FPGAs and ASICs, plus the development of powerful data analytic services.

Earthquake Alert! 6.7 temblor, epicenter 3.8 miles west of Ventura, California--impact will be in eleven minutes--evacuate, evacuate!" While you run to the hall closet to grab your earthquake kit, you shout out: "Alexa, where is my emergency evac location?" Walk north to Wilshire, then take a left on Warner," she responds. As you and your neighbors pour into the building stairwell, you hear audio from a phone: "Google Earth Q estimates substantial potential for structural damage in the West San Fernando Valley and Coastal West Los Angeles to pre-2006 code dwellings and buildings. Most of West LA will experience total loss of power for anywhere from six to twenty-four hours in duration."

With technological advancement, IT technology developers are making efforts to develop software that can ease the physical work life. One such advancement in technology is intelligent automation. The intelligent automation is a combination of automation and artificial intelligence. This new technology has revolutionized the way data is handled and processed. The intelligent automation system determines and synthesizes a massive amount of information, automates the business and operational workflows and adapts it.

While generative models have shown great success in generating high-dimensional samples conditional on low-dimensional descriptors (learning e.g. stroke thickness in MNIST, hair color in CelebA, or speaker identity in Wavenet), their generation out-of-sample poses fundamental problems. The conditional variational autoencoder (CVAE) as a simple conditional generative model does not explicitly relate conditions during training and, hence, has no incentive of learning a compact joint distribution across conditions. We overcome this limitation by matching their distributions using maximum mean discrepancy (MMD) in the decoder layer that follows the bottleneck. This introduces a strong regularization both for reconstructing samples within the same condition and for transforming samples across conditions, resulting in much improved generalization. We refer to the architecture as \emph{transformer} VAE (trVAE). Benchmarking trVAE on high-dimensional image and tabular data, we demonstrate higher robustness and higher accuracy than existing approaches. In particular, we show qualitatively improved predictions for cellular perturbation response to treatment and disease based on high-dimensional single-cell gene expression data, by tackling previously problematic minority classes and multiple conditions. For generic tasks, we improve Pearson correlations of high-dimensional estimated means and variances with their ground truths from 0.89 to 0.97 and 0.75 to 0.87, respectively.

Recent advances in pre-training huge models on large amounts of text through self supervision have obtained state-of-the-art results in various natural language processing tasks. However, these huge and expensive models are difficult to use in practise for downstream tasks. Some recent efforts use knowledge distillation to compress these models. However, we see a gap between the performance of the smaller student models as compared to that of the large teacher. In this work, we leverage large amounts of in-domain unlabeled transfer data in addition to a limited amount of labeled training instances to bridge this gap. We show that simple RNN based student models even with hard distillation can perform at par with the huge teachers given the transfer set. The student performance can be further improved with soft distillation and leveraging teacher intermediate representations. We show that our student models can compress the huge teacher by up to 26x while still matching or even marginally exceeding the teacher performance in low-resource settings with small amount of labeled data.

Recently, there has been significant interest in linear regression in the situation where predictors and responses are not observed in matching pairs corresponding to the same statistical unit as a consequence of separate data collection and uncertainty in data integration. Mismatched pairs can considerably impact the model fit and disrupt the estimation of regression parameters. In this paper, we present a method to adjust for such mismatches under ``partial shuffling" in which a sufficiently large fraction of (predictors, response)-pairs are observed in their correct correspondence. The proposed approach is based on a pseudo-likelihood in which each term takes the form of a two-component mixture density. Expectation-Maximization schemes are proposed for optimization, which (i) scale favorably in the number of samples, and (ii) achieve excellent statistical performance relative to an oracle that has access to the correct pairings as certified by simulations and case studies. In particular, the proposed approach can tolerate considerably larger fraction of mismatches than existing approaches, and enables estimation of the noise level as well as the fraction of mismatches. Inference for the resulting estimator (standard errors, confidence intervals) can be based on established theory for composite likelihood estimation. Along the way, we also propose a statistical test for the presence of mismatches and establish its consistency under suitable conditions.