Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation.

Law makers ask President Obama for more water pumping from Sacramento Delta.

Wildfires are increasingly impacting the environment, human health and safety. Among the top 20 California wildfires, those in 2020-2021 burned more acres than the last century combined. California's 2018 wildfire season caused damages of $148.5 billion. Among millions of impacted people, those living with disabilities (around 15% of the world population) are disproportionately impacted due to inadequate means of alerts. In this project, a multi-modal wildfire prediction and personalized early warning system has been developed based on an advanced machine learning architecture. Sensor data from the Environmental Protection Agency and historical wildfire data from 2012 to 2018 have been compiled to establish a comprehensive wildfire database, the largest of its kind. Next, a novel U-Convolutional-LSTM (Long Short-Term Memory) neural network was designed with a special architecture for extracting key spatial and temporal features from contiguous environmental parameters indicative of impending wildfires. Environmental and meteorological factors were incorporated into the database and classified as leading indicators and trailing indicators, correlated to risks of wildfire conception and propagation respectively. Additionally, geological data was used to provide better wildfire risk assessment. This novel spatio-temporal neural network achieved >97% accuracy vs. around 76% using traditional convolutional neural networks, successfully predicting 2018's five most devastating wildfires 5-14 days in advance. Finally, a personalized early warning system, tailored to individuals with sensory disabilities or respiratory exacerbation conditions, was proposed. This technique would enable fire departments to anticipate and prevent wildfires before they strike and provide early warnings for at-risk individuals for better preparation, thereby saving lives and reducing economic damages.

California is back on the map as a state that's serious about welcoming driverless cars. Truly driverless cars -- vehicles with no human behind the wheel, and perhaps no steering wheel at all -- are headed toward California streets and highways starting in 2018. After months of criticism, state regulators Friday released a proposal for a new set of regulations to govern the testing and deployment of driverless cars on public roadways. They are seeking public comment and expect approval by the end of the year. The regulations lay out "a clear path for future deployment of autonomous vehicles" in California, said Bernard Soriano, deputy director at the Department of Motor Vehicles.

Waymo, the company that was formerly Google's self-driving-car division, on Friday sought a court order to stop Uber from using trade secrets allegedly stolen by a former Waymo employee who took a job with the ride-hailing firm. The preliminary injunction, filed in U.S. District Court, seeks to temporarily prohibit Uber from "accessing, using, imitating, copying, disclosing, or making available to any person or entity Waymo's" trade secrets. The filing requests a hearing occur on April 27 before Judge William H. Alsup. When reached for comment, Uber referred The Times to a statement it had issued a month earlier, describing Waymo's legal actions as a "baseless attempt to slow down a competitor and we look forward to vigorously defending against them in court." Waymo last month sued Uber, alleging that former Waymo employee Anthony Levandowski downloaded more than 14,000 highly confidential and proprietary files shortly before his resignation from the company in January 2016.

The vast majority of members of the American Meteorological Society agree that recent climate change stems at least in part from human causes, and the agreement has been growing significantly in the last five years. According to a new survey of AMS members, 67% say climate change over the last 50 years is mostly to entirely caused by human activity, and more than 4 in 5 (80%) respondents attributed at least some of the climate change to human activity. Roughly 5% said only natural processes were driving climate change. AMS is a broad body that represents expertise including Ph.D. holding scientists and professors (nearly 30%), operational weather forecasters, TV meteorologists, private company professionals, and so on. As such, the 67 to 80 percent number is going to vary from the oft-cited 97% number for climate scientists publishing on the topic or expressing an opinion.