Google has finally revealed a commercial use for DeepMind -- a British artificial intelligence company it acquired for over 600 million in 2014. DeepMind made headlines for beating the best human in the world at the notoriously complex board game Go and it's recently started working with hospitals in the UK on a number of healthcare projects but the startup is yet to make any money for Google, until now. Google announced on Wednesday that it has been using a DeepMind-built AI system to control certain parts of its power-hungry data centers over the last few months as it looks to make its vast server farms more environmentally friendly. Last year, a Greenpeace report predicted that the electricity consumption of data centers is set to account for 12% of global electricity consumption by 2017 and companies like Google, Amazon, Facebook and Apple have some of the biggest data centers in the world. Google said it has been able to reduce the energy consumption of its data center cooling units -- used to stop Google's self-built servers from overheating -- by as much as 40% with the help of a DeepMind AI system.

Google tapped into the superior intelligence of its DeepMind neural network to find ways to vastly reduce the energy it uses in its data centers, which make up 40% of the worldwide Internet. "This will also help other companies who run on Google's cloud to improve their own energy efficiency," Google said in a blog about the achievement. "While Google is only one of many data center operators in the world, many are not powered by renewable energy as we are." Google has set a goal to eventually power its data centers using 100% renewable energy. Today, the company claims, renewable energy is used for 35% of its power needs.

Google tapped into the superior intelligence of its DeepMind neural network to find ways to vastly reduce the energy it uses in its data centers, which make up 40% of the worldwide Internet. "This will also help other companies who run on Google's cloud to improve their own energy efficiency," Google said in a blog about the achievement. "While Google is only one of many data center operators in the world, many are not powered by renewable energy as we are." Google has set a goal to eventually power its data centers using 100% renewable energy. Today, the company claims, renewable energy is used for 35% of its power needs.

Google has created artificial intelligence that's able to save the amount of electricity it uses to power its data centres. Using machine learning developed by the firm's AI research company, DeepMind, it was possible to reduce the energy used for cooling the centres by a staggering 40 per cent. By applying machine learning to its own centres, which power Google Search, Gmail, YouTube and all of Google's services, it was able to improve their efficiency. The algorithms and methods used could also be transferred to air conditioning systems in large manufacturing plants or, on an even larger scale, to reduce wastage in the energy grid. "What we've been trying to do is build a better predictive model that essentially uses less energy to power the cooling system by more accurately predicting when the incoming compute load is likely to land," Mustafa Suleyman, the co-founder of DeepMind told WIRED.

This paper is concerned with the numerical solution of model-based, Bayesian inverse problems. We are particularly interested in cases where the cost of each likelihood evaluation (forward-model call) is expensive and the number of un- known (latent) variables is high. This is the setting in many problems in com- putational physics where forward models with nonlinear PDEs are used and the parameters to be calibrated involve spatio-temporarily varying coefficients, which upon discretization give rise to a high-dimensional vector of unknowns. One of the consequences of the well-documented ill-posedness of inverse prob- lems is the possibility of multiple solutions. While such information is contained in the posterior density in Bayesian formulations, the discovery of a single mode, let alone multiple, is a formidable task. The goal of the present paper is two- fold. On one hand, we propose approximate, adaptive inference strategies using mixture densities to capture multi-modal posteriors, and on the other, to ex- tend our work in [1] with regards to effective dimensionality reduction techniques that reveal low-dimensional subspaces where the posterior variance is mostly concentrated. We validate the model proposed by employing Importance Sam- pling which confirms that the bias introduced is small and can be efficiently corrected if the analyst wishes to do so. We demonstrate the performance of the proposed strategy in nonlinear elastography where the identification of the mechanical properties of biological materials can inform non-invasive, medical di- agnosis. The discovery of multiple modes (solutions) in such problems is critical in achieving the diagnostic objectives.

This paper characterizes hierarchical clustering methods that abide by two previously introduced axioms -- thus, denominated admissible methods -- and proposes tractable algorithms for their implementation. We leverage the fact that, for asymmetric networks, every admissible method must be contained between reciprocal and nonreciprocal clustering, and describe three families of intermediate methods. Grafting methods exchange branches between dendrograms generated by different admissible methods. The convex combination family combines admissible methods through a convex operation in the space of dendrograms, and thirdly, the semi-reciprocal family clusters nodes that are related by strong cyclic influences in the network. Algorithms for the computation of hierarchical clusters generated by reciprocal and nonreciprocal clustering as well as the grafting, convex combination, and semi-reciprocal families are derived using matrix operations in a dioid algebra. Finally, the introduced clustering methods and algorithms are exemplified through their application to a network describing the interrelation between sectors of the United States (U.S.) economy.

Distributed learning is the problem of inferring a function in the case where training data is distributed among multiple geographically separated sources. Particularly, the focus is on designing learning strategies with low computational requirements, in which communication is restricted only to neighboring agents, with no reliance on a centralized authority. In this thesis, we analyze multiple distributed protocols for a large number of neural network architectures. The first part of the thesis is devoted to a definition of the problem, followed by an extensive overview of the state-of-the-art. Next, we introduce different strategies for a relatively simple class of single layer neural networks, where a linear output layer is preceded by a nonlinear layer, whose weights are stochastically assigned in the beginning of the learning process. We consider both batch and sequential learning, with horizontally and vertically partitioned data. In the third part, we consider instead the more complex problem of semi-supervised distributed learning, where each agent is provided with an additional set of unlabeled training samples. We propose two different algorithms based on diffusion processes for linear support vector machines and kernel ridge regression. Subsequently, the fourth part extends the discussion to learning with time-varying data (e.g. time-series) using recurrent neural networks. We consider two different families of networks, namely echo state networks (extending the algorithms introduced in the second part), and spline adaptive filters. Overall, the algorithms presented throughout the thesis cover a wide range of possible practical applications, and lead the way to numerous future extensions, which are briefly summarized in the conclusive chapter.

Google said it now gets 3.5 times as much computing power out of the same amount of energy as it did five years ago thanks to custom-built servers, more efficient cooling systems that use outside air, and investment in green energy. The company wants to cap its increase in energy use at four per cent a year between 2014 and 2020 even as data use grows at a faster rate. It also plans to be 100 per cent powered by renewable energy. But it hasn't said when it will reach that goal, or how much of its power currently comes from renewable sources. The data centre algorithm can eventually be used to improve efficiency in other areas, according to Google, including getting more energy from the same amount of input at power plants, and reducing energy and water usage in semiconductor manufacturing.

The Intergovernmental Panel on Climate Change (IPCC) highlights six main lines of evidence for climate change. First, we have tracked (see chart) the unprecedented recent increase in the amount of atmospheric carbon dioxide and other greenhouse gases since the beginning of the industrial revolution. By burning coal, oil, and natural gas, we accelerate the process, releasing vast amounts of carbon (carbon that took millions of years to accumulate) into the atmosphere every year.

Google's grand experiment in using artificial intelligence to save power is paying off. The search firm's Demis Hassabis tells Bloomberg that the DeepMind AI has cut electricity use at Google data centers by "several percentage points" thanks to its extra-efficient use of equipment, such as cooling systems and windows. It's not certain just how much energy the smart code is saving, but Google used slightly over 4.4 gigawatt-hours in 2014 alone -- even a small dent in that consumption could easily save hundreds of millions of dollars. That pricey DeepMind acquisition is likely paying for itself.