We consider the problem where $n$ clients transmit $d$-dimensional real-valued vectors using $d(1+o(1))$ bits each, in a manner that allows the receiver to approximately reconstruct their mean. Such compression problems naturally arise in distributed and federated learning. We provide novel mathematical results and derive computationally efficient algorithms that are more accurate than previous compression techniques. We evaluate our methods on a collection of distributed and federated learning tasks, using a variety of datasets, and show a consistent improvement over the state of the art.

Google's Latest AI Ransomware Defense Only Goes So Far Google has launched a new AI-based protection in Drive for desktop that can shut down an attack before it spreads--but its benefits have their limits. Ransomware attacks have loomed for years as an urgent digital threat with no easy solution --especially as they have evolved to include data grab-and-leak attacks that may not even involve data-encrypting malware at all. Traditional ransomware that locks up files and systems is still rampant, though, and Google on Tuesday launched a new defense for its Google Drive for desktop apps that aims to quickly detect ransomware activity and halt cloud syncing before an infection can spread. While antivirus scanners monitor for signs of malware across a system, the new ransomware protections in Drive for desktop are meant to act as an additional line of defense. The detection capability is built on an AI model that Google trained using millions of real victims' files that had been encrypted with various strains of ransomware.

This let Flakey drive along hallways with no dead reckoning or planning whatsoever. It seemed miraculous at the time; a situated automaton that knew things without needing any models. However, I thought of it as (sensor-driven) feedback control, versus (plan driven, eyes shut) feed-forward control. I then used Mike Georgeff's procedural reasoning system (PRS) to make Flakey not only drive but navigate an office building. In some respects this project succeeded: the robot's "domain knowledge" was nothing more than a static connection graph--no distances to drive, no widths of halls or doorways, no a priori obstacles--such information was acquired en route from sensory input.

The primary revolution in automated planning in the last decade has been the very impressive scaleup in planner performance. A large part of the credit for this can be attributed squarely to the invention and deployment of powerful reachability heuristics. Most, if not all, modern reachability heuristics are based on a remarkably extensible data structure called the planning graph, which made its debut as a bit player in the success of GraphPlan, but quickly grew in prominence to occupy the center stage. Planning graphs are a cheap means to obtain informative look-ahead heuristics for search and have become ubiquitous in state-of-the-art heuristic search planners. We present the foundations of planning graph heuristics in classical planning and explain how their flexibility lets them adapt to more expressive scenarios that consider action costs, goal utility, numeric resources, time, and uncertainty. Considerable work has been done in the last 40 years on modeling a wide variety of ...

Are you a veteran of L.A.'s current dating scene? Even in the New Los Angeles, with Lyft and Uber giving us cheaper rides and two-thirds of voters passing Measure M, dating without a car is still playing the game with a serious handicap. "I don't mind that you don't drive," a woman I'd been dating for six months told me last year as she drove us to dinner at Broken Spanish for my birthday. L.A. Affairs chronicles the current dating scene in and around Los Angeles.

Simply put, machine learning and AI, in general, will become commonplace in our lives because we need them to be. Alongside development of the capability to process massive amounts of data in innovative ways, there exists another technical revolution whose time has also most definitely come -- the internet of things (IoT). If AI offers the promise of processing immense quantities of data in ways that we can't, then IoT provides the very tangible mechanism for generating that raw data in ways we might not expect. Perhaps more telling, there is already an emerging trend of AI development "following the data" in order to accelerate the capability to deliver human-machine interactions and insight based on the availability of more of those very same interactions.

According to a global survey of 260 large organizations conducted by market research firm Vanson Bourne on behalf of Teradata, a data and analytics company, 80 percent of enterprises are investing in AI and one in three "believe their company will need to invest more over the next 36 months to keep pace with competitors." In general, enterprise AI includes things like machine and deep learning, voice recognition and response, Robotic Process Automation (RPA), automated communications and reporting, predictive analytics, and recommendation engines. The report reaffirms the findings of a similar study from Narrative Science last year, which found that 59 percent of respondents saw the shortage of data science talent as the primary barrier to getting the full value out of emerging Big Data and AI technologies. But it's going to happen quickly: Currently, most businesses are using existing technology leaders, such as CTOs, to drive AI deployment and strategy.

But while automakers are still dropping level 2 and sometimes level 3 vehicles into the market, NVIDIA has announced its first AI computer, the NVIDIA Drive PX Pegasus that it says is capable of level 5 autonomy. The computing needed to power a self-driving car's AI and data crunching not to mention the huge amounts of data coming from potentially dozens of cameras, LiDAR sensors, short and long-range radar is staggering and usually means there's a small server room stored in the trunk. The new NVIDIA Drive PX Pegasus AI computer is the size of a license plate and uses far less power than the current model. The delivery service is looking to deploy a pilot fleet with the current Drive PX in 2018.