The radioactive'miracle water' that killed its believers In the 1920s, Radithor promised to cure everything from wrinkles to leukemia, but its unintended results were deadly. While 1920s soda shops offered a plethora of sweet treats, nearby pharmacies served their own tinctures--like Radithor, certified radioactive water. Breakthroughs, discoveries, and DIY tips sent every weekday. William Bailey promised to cure anything that ailed you. " Just a tiny bottle of apparently lifeless, colorless, and tasteless water " was, he advertised in a 1929 pamphlet for his product, Radithor, "the greatest therapeutic force known to mankind."

Before autonomous systems can be deployed in safety-critical applications, we must be able to understand and verify the safety of these systems. For cases where the risk or cost of real-world testing is prohibitive, we propose a simulation-based framework for a) predicting ways in which an autonomous system is likely to fail and b) automatically adjusting the system's design and control policy to preemptively mitigate those failures. Existing tools for failure prediction struggle to search over high-dimensional environmental parameters, cannot efficiently handle end-to-end testing for systems with vision in the loop, and provide little guidance on how to mitigate failures once they are discovered. We approach this problem through the lens of approximate Bayesian inference and use differentiable simulation and rendering for efficient failure case prediction and repair. For cases where a differentiable simulator is not available, we provide a gradient-free version of our algorithm, and we include a theoretical and empirical evaluation of the trade-offs between gradient-based and gradient-free methods. We apply our approach on a range of robotics and control problems, including optimizing search patterns for robot swarms, UAV formation control, and robust network control. Compared to optimization-based falsification methods, our method predicts a more diverse, representative set of failure modes, and we find that our use of differentiable simulation yields solutions that have up to 10x lower cost and requires up to 2x fewer iterations to converge relative to gradient-free techniques. In hardware experiments, we find that repairing control policies using our method leads to a 5x robustness improvement. Accompanying code and video can be found at https://mit-realm.github.io/radium/

Today, Radium, a startup that aims to use artificial intelligence and machine learning to extract more computing power from cloud hardware, announced it was leaving stealth mode and deploying its solutions to cloud datacenters run by Cyxtera in Toronto, the New York and New Jersey metro area, and Silicon Valley. The main product, called Launchpad, lets users start and shut down projects on bare metal machines, eliminating the extra layers of hypervisors and virtualization software. Radium offered benchmark tests on machine learning jobs that showed speed increases ranging from 30% and 140%. "Our initial testing shows that bare metal servers offer a good cloud computing platform for the high-performance deep learning and inference workloads required for these types of applications," said Srinivasa Narasimhan, a professor at Carnegie Mellon's School of Computer Science, who has been working with the company to test its product. Many cloud products rely heavily on virtualization software layers, or "hypervisors," that allow one physical machine to simulate a variety of smaller machines that appear independent to users.

This year brought no shortage of great science-themed books. Spurred by rapid advances in biotech, the writer Carl Zimmer spun a personal tale around the emerging science of heredity. Investigative reporter John Carreyrou exposed the rotten business at the heart of Theranos, the blood-testing startup built on air. Our past also proved bountiful, with books on that time we made teenage girls glow until their bones rotted (The Radium Girls), and when competing visionaries dueled over how to steward our one and only world (The Wizard and the Profit). If that all seems a bit much, we've got an escape hatch: psychedelics. Lots of them, as recounted by Michael Pollan. But those are just a few of the superb tomes to emerge in 2018.