Beyond Limits, a Glendale, California-based company developing explainable AI for a range of industries, today announced a $25 million strategic partnership with XCELL to build what it’s calling the world’s first AI-controlled power plant. Beyond Limits CEO AJ Abdallat said it’ll be designed from the ground up and from scratch, and that it’ll embed intelligence into the “entire operation” to make the plant “safer” and maximize efficiency, productivity, and environmental friendliness.
Above: Beyond Limits CEO AJ Abdallat onstage at VentureBeat Transform 2019. Photo Credit: VentureBeat
“We’re excited about it,” said Abdallat at VentureBeat’s Transform 2019 conference in San Francisco. “State of the art today is predicting whether a device is going to fail. We do all of that, and that’s wonderful, but [we also have a] layer on top [that allows us to] align with the objectives of the company managing [the facility] so they can build a system that oversees everything going on.”
It’s an approach akin to that taken by Foxconn-backed Carbon Relay, which aims to cut datacenter carbon emissions with tools that draw on thousands of sensors to make predictions about electrical usage. Similarly, Google parent company Alphabet’s DeepMind said in February that it’s using AI to forecast the performance of wind turbines 36 hours in advance, and last August, DeepMind revealed that it’s turned over management of one of Google’s datacenter cooling controls to an AI-powered recommender system.
“When you make a change in one portion of [a facility], whether you like it or not, you’re going to change things throughout the system. You can’t ever hit pause to try and figure out what’s going on,” said Beyond Limits technical head of industrial IoT Kristin Lennox, who demoed the company’s system onstage as applied to an oil refinery. “The stakes are high. If you make a bad decision, there are potentially dire consequences in terms of health and safety.”
That’s why Beyond Limits champions what it calls “cognitive AI,” a paradigm that entails ingesting data from monitoring devices to create a picture of what’s going on at a given asset. A machine learning system detects anomalies in part with the aid of a second layer — a “cognitive” layer — that includes facility goals, and it provides a full trace of all data and all steps used to arrive at its conclusions. Simultaneously, the system evaluates situations to identify where it’s not achieving baseline goals.
“In addition to constantly evaluating operations and trying to figure out how we make those as good as they can be, we’re simultaneously … reporting back to engineers the factors that are driving [shortcomings] and the ways they [can be addressed],” said Lennox.
Beyond Limits was co-founded in 2014 by Abdallat and Mark James as a full-stack engineering company creating AI solutions deployable via the cloud, on-premises, or embedded in devices at the edge. In 2012, the California Institute of Technology granted the company a license to update and commercialize 40 of the AI programs developed through NASA’s Deep Space program and operated out of Caltech’s Jet Propulsion Laboratory (JPL).
One of the aforementioned programs is Spacecraft Health Inference Engine (SHINE), a software-defined knowledge-based system designed to be efficient enough to operate in real-time environments and used by non-LISP apps written in programming languages like C and C++. It’s used at JPL, where it’s been applied to AI research and specialized tools running on distributed systems.
“When you’re operating in space, you’re operating in a very harsh and dynamic environment,” explained Abdallat. “You don’t have access to the computing power — you don’t have access to Amazon’s, Microsoft’s, or Google’s amazing cloud. And you’re going to have gaps in the communication — you’re going to have issues with the data. As a result, [NASA] created a world-class organization to focus on how to … allow these missions to [succeed] on [long] journeys.”
Today, Beyond Limits’ suite includes a management adviser for energy operators that integrates with platforms for physics-based, geology, and policy elements, and health care products that combine conventional AI techniques with symbolic reasoning to analyze patient data, lab results, chart notes, and drug interactions and proactively alert medical staff about deteriorating patients. Meanwhile, in the automotive sector, the company’s systems can predict if drivers have become distracted from safe driving and intervene.
Read more on VentureBeat.com