Recognizing symmetries in data allows for significant boosts in neural network training, which is especially important where training data are limited. In many cases, however, the exact underlying symmetry is present only in an idealized dataset, and is broken in actual data, due to asymmetries in the detector, or varying response resolution as a function of particle momentum. Standard approaches, such as data augmentation or equivariant networks fail to represent the nature of the full, broken symmetry, effectively overconstraining the response of the neural network. We propose a learning model which balances the generality and asymptotic performance of unconstrained networks with the rapid learning of constrained networks. This is achieved through a dual-subnet structure, where one network is constrained by the symmetry and the other is not, along with a learned symmetry factor. In a simplified toy example that demonstrates violation of Lorentz invariance, our model learns as rapidly as symmetry-constrained networks but escapes its performance limitations.

Large Language Models (LLMs) exhibit impressive results across a wide range of natural language processing (NLP) tasks, yet they can often produce factually incorrect outputs. This paper introduces a simple but effective low-latency post-correction method, \textbf{Retrieval Augmented Correction (RAC)}, aimed at enhancing the factual performance of LLMs without requiring additional fine-tuning. Our method is general and can be used with any instruction-tuned LLM, and has greatly reduced latency compared to prior approaches. RAC decomposes the LLM's output into atomic facts and applies a fine-grained verification and correction process with retrieved content to verify and correct the LLM-generated output. Our extensive experiments show that RAC yields up to 30\% improvements over state-of-the-art baselines across two popular factuality evaluation datasets, validating its efficacy and robustness in both with and without the integration of Retrieval-Augmented Generation (RAG) across different LLMs.\footnote{Our code is at \url{https://github.com/jlab-nlp/Retrieval-Augmented-Correction}}

Farmers Insurance (Woodland Hills, Calif.) has announced plans to use a digitally controlled mobile robot to assist with in-field catastrophe claims handling and non-catastrophe property inspections, with the aim of helping to improve the safety and efficiency of both while becoming one of the first national property/casualty insurers to deploy a robotic quadruped. Created by Boston Dynamics (Waltham, Mass.) and customized for Farmers, the robot--named "Spot"--will be used by Farmers claims personnel as early as fall 2021 to help assess damage from catastrophes such as hurricanes, tornadoes, earthquakes, and wildfires. Farmers reports that Spot will be equipped with a variety of sensors and cameras, including a 360-degree camera and site documentation software to help reduce the time required to capture data and augment the in-field claims review process. The robot may also help Farmers claims adjusters collect critical data and assist with claims handling optimization to serve impacted customers more efficiently, according to a statement from the insurer. In addition to utilizing Spot for in-field claims use, Farmers says it will explore applications that could help first-responder organizations during scenarios such as post-event search and rescue operations, accessing areas to assess danger for first responders or others, and/or pre-inspections to assess safety for anyone in the general vicinity.

Patrick Kelly remembers the pitch meeting vividly. The room full of developers and Activision executives had convened at Infinity Ward's offices in Woodland Hills, California, in early 2018. It was time for Kelly and his longtime colleague Dave Stohl, who together serve as co-studio heads for Infinity Ward, to pitch their big idea. The project was code-named "Magma." And the plan was to create the biggest ever battle royale, one tied to the world of the studio's planned 2019 release, "Call of Duty: Modern Warfare." Upon completion, the project would be re-christened as "Warzone." "Okay, so here's the thing," Kelly said, reenacting his pitch to the room.

There's plenty of competition: VirtualAPT, based in Brooklyn, has robots that glide through homes and provide immersive virtual reality tours; REX, a brokerage in Woodland Hills, Calif., has an AI-trained robot to answer potential buyers' questions at open houses; RealFriend and OjoLabs have AI-powered chatbots that mimic human conversation while providing deeply personalized home listings and buying advice. In Zenny's case, the robot is powered remotely by the real estate broker or property manager who is handling the showing from afar. It is also equipped with sensors to keep it from running into walls or people. In addition to Zenny, Zenplace's platform includes a full suite of rental management solutions, including tenant screening, electronic lockboxes for on-demand property viewings, and a secure online portal for rent payment. The company charges a $599 flat fee for some properties, and $99 a month for others. VirtualAPT's robots, which roll through homes capturing 360-degree videos in 4K resolution, provide ultra-crisp, high-quality images.

Amazon is soft-launching a new supermarket Thursday, with a shopping cart that tallies up items as it enters the basket and enables instant checkout, and Alexa stations throughout the store you can ask questions. Amazon's first Fresh store is a 35,000 square foot traditional supermarket, opening in a strip mall in Woodland Hills, California, next to a See's Candy. Woodland Hills is a Los Angeles suburb in the heart of the San Fernando Valley. Customers can enter the store the traditional way, but if they want to use the cart, they need to open up their Amazon app and swipe it. Jeff Helbling, Amazon's vice-president of the Fresh Stores, says the shopping cart uses "a combination of computer vision algorithms and sensor fusion," within the cart to identify and tally up the items.

Amazon (AMZN) has introduced shopping carts that make it faster and more convenient to shop by automatically tracking the items put in the cart, enabling consumers to eliminate the checkout line. The new Dash Carts will first be featured at Amazon's Woodland Hills, California, grocery store, set to open this year. To use the Dash Carts, shoppers will need to have an Amazon account and a smartphone. Shoppers simply scan a QR code located within the Amazon app to begin loading items into the cart. The Smart Cart is fitted with computer vision algorithms and sensor fusion to recognize merchandise that is put into the cart.

Amazon has created a smart shopping cart that knows what you're selecting and can charge you for it without a cashier. It's called the "Amazon Dash Cart" and the idea is to make "a quick grocery trip even quicker by allowing you to skip the checkout line," the e-commerce giant said in a post on Tuesday. The Dash Cart largely looks like a typical shopping cart, only bulkier at the bottom. Patrons will use their Amazon account's unique QR code to sign-in to the cart, according to a video the company posted on its website. Then you add your shopping bags to the cart and proceed to shop around the store.

Report explores how organizations using AI are making it work, what technologies they're using, and what best practices can maximize an organization's success with AI and machine learning Seattle, WA, Oct. 18, 2019 (GLOBE NEWSWIRE) -- TDWI Research has released its newest Best Practices Report, Driving Digital Transformation using AI and Machine Learning. This original, survey-based report looks at the many dimensions of artificial intelligence (AI) and machine learning (ML) so data professionals and their business counterparts can understand the benefits of the technology, how it's used (and by whom), and how enterprises are achieving success with it. The author of the report, Fern Halper, is vice president and senior director of TDWI Research for advanced analytics. She explains that organizations are embracing AI and ML to gain better insights, make better decisions, and improve competitive advantage. "In fact," she writes, "AI is at the heart of the digital revolution around analytics occurring today. AI promises to help organizations improve their operations and processes and to drive new revenue opportunities."

The finalists for the 2018 Nebula Awards, considered some of the most prestigious in science fiction and fantasy, were announced on Wednesday, with novelists Tomi Adeyemi and Justina Ireland, filmmaker Boots Riley and musician Janelle Monáe among the nominees. Adeyemi and Ireland were nominated in the young adult category, Adeyemi for "Children of Blood and Bone" and Ireland for "Dread Nation." Both books gained critical acclaim upon their release, and Adeyemi's book was the inaugural pick for Jimmy Fallon's "Tonight Show" book club. The two authors are joined in the category by Roshani Chokshi's "Aru Shah and the End of Time," A.K. DuBoff's "A Light in the Dark," Rachel Hartman's "Tess of the Road" and Henry Lien's "Peasprout Chen: Future Legend of Skate and Sword." Finalists in the novel category included Mary Robinette Kowal's "The Calculating Stars" and R.F.