Google.org is today releasing a report that details how social impact startups, nonprofits, government policymakers, and academics can use machine learning to address some of humanity's biggest problems. The assessment is the result of analysis of more than 2,600 applications -- from 119 countries -- to the $25 million Google AI Impact Challenge, a global open call for projects that aim to use AI to serve humanity. Google used AI to arrive at its conclusions about how human-centered organizations can better use AI -- tapping natural language processing and clustering analysis techniques to review applications and internal assessment of applications. The AI analysis was supplemented by interviews with impact challenge applicants and recipients. Today's assessment of how civil society can improve AI implementations appears to be the first such report from a tech giant.

Based on some of our recent research and learnings, here are some of the most common use cases of Artificial Intelligence and Machine Learning techniques used in finance. This link is a good introduction of machine learning. This article is good read for both technical and non technical people. Fraud detection is topic that is applicable to many industries like Banking, Finance, Insurance, Government policies etc. Recent rise in digitization fraud is becoming the major problem that finance and banking institutions faces.

AI and machine learning is expected to play a big role in radiology and medical imaging as algorithms speed up the review process.

Fox News Flash top headlines for Sept. 12 are here. Check out what's clicking on Foxnews.com California could soon become the largest state to ban the use of facial recognition technology in law enforcement body cameras, a significant milestone in the regulation of the burgeoning technology. The State Assembly on Thursday passed AB 1215, a bill that would impose a three-year moratorium on the technology, garnering praise from privacy and civil liberties advocates. The legislation now heads to Gov. Gavin Newsom's desk.

In the 1970s, the U.S. Department of Energy poured money into making practical a miraculous technology: the ability to convert sunlight into electricity. Solar energy was a pipe dream, far too expensive and unreliable to be considered a practical power source. But yesterday's moon shot is today's reality. The expense of solar power has fallen more quickly than expected, with installations costing about 80% less today than a decade ago. Alternative energy (like wind and solar) is now often cheaper than conventional energy (like coal and gas).

Iowa State University scientists are working toward a future in which farmers can use unmanned aircraft to spot, and even predict, disease and stress in their crops. Their vision relies on machine learning, an automated process in which technology can help farmers respond to plant stress more efficiently. Arti Singh, an adjunct assistant professor of agronomy, is leading a multi-disciplinary research team that recently received a three-year, $499,845 grant from the U.S Department of Agriculture's National Institute of Food and Agriculture to develop machine learning technology that could automate the ability of farmers to diagnose a range of major stresses in soybeans. The technology under development would make use of cameras attached to unmanned aerial vehicles, or UAVs, to gather birds-eye images of soybean fields. A computer application would automatically analyze the images and alert the farmer of trouble spots.

The Air Force released the Artificial Intelligence Strategy Sept. 12, highlighting the importance of artificial intelligence capabilities to 21st century missions. The strategy provides definition, context and purpose for artificial intelligence in the Air Force, and is the service's annex to the Department of Defense Artificial Intelligence Strategy. "The Air Force is charged to provide the nation with Air and Space Superiority, Global Strike, Rapid Global Mobility, Intelligence, Surveillance and Reconnaissance and Command Control," said Acting Secretary of the Air Force Matthew Donovan and Air Force Chief of Staff General David Goldfein in the dual-signed annex. "Al is a capability that will underpin our ability to compete, deter and win across all five of these diverse missions. It is crucial to fielding tomorrow's Air Force faster and smarter, executing multi-domain operations in the high-end fight, confronting threats below the level of open conflict and partnering with our allies around the globe." The strategy serves as the framework for aligning Air Force efforts with the National Defense Strategy and the Department of Defense Artificial Intelligence Strategy as executed by the Joint Artificial Intelligence Center.

Data fields sampled on irregularly spaced points arise in many applications in the sciences and engineering. For regular grids, Convolutional Neural Networks (CNNs) have been successfully used to gaining benefits from weight sharing and invariances. We generalize CNNs by introducing methods for data on unstructured point clouds based on Generalized Moving Least Squares (GMLS). GMLS is a non-parametric technique for estimating linear bounded functionals from scattered data, and has recently been used in the literature for solving partial differential equations. By parameterizing the GMLS estimator, we obtain learning methods for operators with unstructured stencils. In GMLS-Nets the necessary calculations are local, readily parallelizable, and the estimator is supported by a rigorous approximation theory. We show how the framework may be used for unstructured physical data sets to perform functional regression to identify associated differential operators and to regress quantities of interest. The results suggest the architectures to be an attractive foundation for data-driven model development in scientific machine learning applications.

We propose a scheme for defending against adversarial attacks by suppressing the largest eigenvalue of the Fisher information matrix (FIM). Our starting point is one explanation on the rationale of adversarial examples. Based on the idea of the difference between a benign sample and its adversarial example is measured by the Euclidean norm, while the difference between their classification probability densities at the last (softmax) layer of the network could be measured by the Kullback-Leibler (KL) divergence, the explanation shows that the output difference is a quadratic form of the input difference. If the eigenvalue of this quadratic form (a.k.a. FIM) is large, the output difference becomes large even when the input difference is small, which explains the adversarial phenomenon. This makes the adversarial defense possible by controlling the eigenvalues of the FIM. Our solution is adding one term representing the trace of the FIM to the loss function of the original network, as the largest eigenvalue is bounded by the trace. Our defensive scheme is verified by experiments using a variety of common attacking methods on typical deep neural networks, e.g. LeNet, VGG and ResNet, with datasets MNIST, CIFAR-10, and German Traffic Sign Recognition Benchmark (GTSRB). Our new network, after adopting the novel loss function and retraining, has an effective and robust defensive capability, as it decreases the fooling ratio of the generated adversarial examples, and remains the classification accuracy of the original network.

Entity resolution (ER) (record linkage or de-duplication) is the process of merging together noisy databases, often in the absence of a unique identifier. A major advancement in ER methodology has been the application of Bayesian generative models. Such models provide a natural framework for clustering records to unobserved (latent) entities, while providing exact uncertainty quantification and tight performance bounds. Despite these advancements, existing models do not scale to realistically-sized databases (larger than 1000 records) and they do not incorporate probabilistic blocking. In this paper, we propose "distributed Bayesian linkage" or d-blink -- the first scalable and distributed end-to-end Bayesian model for ER, which propagates uncertainty in blocking, matching and merging. We make several novel contributions, including: (i) incorporating probabilistic blocking directly into the model through auxiliary partitions; (ii) support for missing values; (iii) a partially-collapsed Gibbs sampler; and (iv) a novel perturbation sampling algorithm (leveraging the Vose-Alias method) that enables fast updates of the entity attributes. Finally, we conduct experiments on five data sets which show that d-blink can achieve significant efficiency gains -- in excess of 300$\times$ -- when compared to existing non-distributed methods.