Plastic waste is a significant environmental pollutant that is difficult to monitor. We created a system of neural networks to analyze spectral, spatial, and temporal components of Sentinel-2 satellite data to identify terrestrial aggregations of waste. The system works at continental scale. We evaluated performance in Indonesia and detected 374 waste aggregations, more than double the number of sites found in public databases. The same system deployed across twelve countries in Southeast Asia identifies 996 subsequently confirmed waste sites. For each detected site, we algorithmically monitor waste site footprints through time and cross-reference other datasets to generate physical and social metadata. 19% of detected waste sites are located within 200 m of a waterway. Numerous sites sit directly on riverbanks, with high risk of ocean leakage.

We discuss an approach to probabilistic forecasting based on two chained machine-learning steps: a dimensional reduction step that learns a reduction map of predictor information to a low-dimensional space in a manner designed to preserve information about forecast quantities; and a density estimation step that uses the probabilistic machine learning technique of normalizing flows to compute the joint probability density of reduced predictors and forecast quantities. This joint density is then renormalized to produce the conditional forecast distribution. In this method, probabilistic calibration testing plays the role of a regularization procedure, preventing overfitting in the second step, while effective dimensional reduction from the first step is the source of forecast sharpness. We verify the method using a 22-year 1-hour cadence time series of Weather Research and Forecasting (WRF) simulation data of surface wind on a grid.

At around the age of 12, David Mayman tried to build a helicopter out of fence posts and an old lawn mower. Needless to say, it did not go well. His contraption didn't fly and he was made to fix the fence. "I was brought up in a way that I guess challenged me scientifically... I was always told that nothing's impossible," he says.

We consider the problem of learning the dynamics of autonomous linear systems (i.e., systems that are not affected by external control inputs) from observations of multiple trajectories of those systems, with finite sample guarantees. Existing results on learning rate and consistency of autonomous linear system identification rely on observations of steady state behaviors from a single long trajectory, and are not applicable to unstable systems. In contrast, we consider the scenario of learning system dynamics based on multiple short trajectories, where there are no easily observed steady state behaviors. We provide a finite sample analysis, which shows that the dynamics can be learned at a rate $\mathcal{O}(\frac{1}{\sqrt{N}})$ for both stable and unstable systems, where $N$ is the number of trajectories, when the initial state of the system has zero mean (which is a common assumption in the existing literature). We further generalize our result to the case where the initial state has non-zero mean. We show that one can adjust the length of the trajectories to achieve a learning rate of $\mathcal{O}(\sqrt{\frac{\log{N}}{N})}$ for strictly stable systems and a learning rate of $\mathcal{O}(\frac{(\log{N})^d}{\sqrt{N}})$ for marginally stable systems, where $d$ is some constant.

Which three-word phrase should always be spoken cautiously? All of them, actually, but that last one -- depending on your choice of ride, a full tank of gas can now cost you within fumes-sniffing distance of a hundred bucks. How did it come to this -- again? Los Angeles is a complex place. In this weekly feature, Patt Morrison is explaining how it works, its history and its culture.

As companies attempt to take sustainability to the next level and gain a more complete view of their greenhouse gas emissions, there's a growing need to quantify results and track progress. "If you can't measure it, you can't manage it," says Autumn Stanish, associate principal analyst at Gartner, Inc. "In order to take initiatives to the next level -- particularly as organizations look to expand beyond Scope 1 and Scope 2 tracking -- there's a need for more advanced and granular measurement tools." Boston Consulting Group (BCG) reports that while 85% of companies are interested in reducing their emissions, only 9% of companies measure their total emissions comprehensively. Worse, only 11% have reduced their emissions in line with their goals over the last five years. How can companies get a better handle on their carbon footprint?

There are several risks involved in dealing with Artificial Intelligence (AI). In the globalized world, however, such methodology choices can eventually snowball into much greater economic risks. Let me make a case for an AI Risk Manager in organizations, preferably sitting in the Risk Management Department (RMD) if not compliance. The series of recent AI mishaps have further ignited the debate. A person from Michigan sued the Detroit police after being falsely arrested and falsely identified as a shoplifting suspect by the department's facial recognition software.

Making good use of data and analytics will not be done in any single bold move but through multiple coordinated actions. Despite the recent and significant advances in machine intelligence, the full scale of the opportunity is just beginning to unfold. But why are some companies doing better than others? How do companies identify where to get started based on their digital journeys? In this episode of McKinsey Talks Operations, Bruce Lawler, managing director for the Massachusetts Institute of Technology's (MIT) Machine Intelligence for Manufacturing and Operations (MIMO) program, and Vijay D'Silva, senior partner emeritus at McKinsey, speak with McKinsey's Daphne Luchtenberg about how companies across industries and sizes can learn from leaders and integrate analytics and data to improve their operations. The following is an edited version of their conversation. Daphne Luchtenberg: Earlier this year, McKinsey and MIT's Machine Intelligence for Manufacturing and Operations studied 100 companies and sectors from automotive to mining. To discuss this and more, I'm joined by the authors, Vijay D'Silva, senior partner emeritus at McKinsey, and Bruce Lawler, managing director for MIT's MIMO. Let's start with the why.

Mentions of artificial intelligence within the filings of companies in the power industry were 268% higher between October 2020 and September 2021 …

We propose a unified data-driven reduced order model (ROM) that bridges the performance gap between linear and nonlinear manifold approaches. Deep learning ROM (DL-ROM) using deep-convolutional autoencoders (DC-AE) has been shown to capture nonlinear solution manifolds but fails to perform adequately when linear subspace approaches such as proper orthogonal decomposition (POD) would be optimal. Besides, most DL-ROM models rely on convolutional layers, which might limit its application to only a structured mesh. The proposed framework in this study relies on the combination of an autoencoder (AE) and Barlow Twins (BT) self-supervised learning, where BT maximizes the information content of the embedding with the latent space through a joint embedding architecture. Through a series of benchmark problems of natural convection in porous media, BT-AE performs better than the previous DL-ROM framework by providing comparable results to POD-based approaches for problems where the solution lies within a linear subspace as well as DL-ROM autoencoder-based techniques where the solution lies on a nonlinear manifold; consequently, bridges the gap between linear and nonlinear reduced manifolds. We illustrate that a proficient construction of the latent space is key to achieving these results, enabling us to map these latent spaces using regression models. The proposed framework achieves a relative error of 2% on average and 12% in the worst-case scenario (i.e., the training data is small, but the parameter space is large.).