Security Startup SparkCognition Uses AI to Protect Connected Devices
When trying to solve a problem, people often disregard possible solutions because they don’t seem to make sense. But in doing so, their brains may prevent them from solving certain problems—in particular, those that involve vast amounts of data. That’s part of the premise on which SparkCognition’s business is built.
“My mind, because of common sense, will automatically prune many possibilities,” says Amir Husain, CEO and founder of Austin, TX-based SparkCognition. “In some cases, it works out really well. In other cases, it doesn’t because solutions might elude common sense, solutions might be pruned out by common sense.”
Machines that learn—using artificial intelligence techniques—don’t have the same trouble. “Machine thought is very different,” Husain says. “It can afford to look at a massive set of possibilities. Because it can afford to, it can by definition find things that are counter-intuitive because it’s not a slave to intuition.”
That view led Husain to write the algorithms from which SparkCognition’s artificial intelligence software was built. The company’s software in essence identifies safety and security problems and predicts when others will occur—and seeks to do it for businesses more quickly and efficiently than a human could.
SparkCognition specifically focuses on the equipment of energy companies, as well as governments with Internet-connected physical assets that are potentially vulnerable to cyber attack.
The company, which officially launched its business in early 2014, was named to a list of Austin’s hottest startups in May by the Austin Chamber of Commerce. It has venture backing from Michael Dell’s private investment firm, MSD Capital, as well as investors including the Barragán family out of Mexico, which runs Arca Continental, Husain says. He declined to disclose the amount invested, saying only that it was multiple millions of dollars.
SparkCognition has annual revenue in the millions, too, Husain says. (He also declined to disclose exactly how much.) The company’s fees depend on the number of assets it is monitoring, and the type of customer. SparkCognition has worked with industrial companies that are “million dollar-plus” customers, Husain said, declining to provide more specifics on pricing. The 20 companies it has worked with in its first year receive dashboard and analytics services based on the data SparkCognition’s software studies, Husain says.
One customer that owns railcars has access to a dashboard with each car’s details logged: how old the cars are, how long since they’ve been serviced, what type of service they’ve had, and dozens of other data points.
SparkCognition’s software uses that information to predict when part of the railcar, like a wheel or axle, may start malfunctioning—by going out of alignment, for example, which could cause damage to a rail line and incur a fine from the railroad operator, says John King, an account manager for SparkCognition. The railcar owner can also use SparkCognition’s dashboard service to plan out deliveries—selecting or avoiding rail cars it wants to send on trips based on the AI analytics. That could help it avoid picking a car that will break down partway through the trip.
Each customer dashboard is customized, though the company doesn’t require customers to use the dashboard, Husain says. It also offers a licensing model to companies that want to run SparkCognition’s software using their own systems, he says.
The company has developed at least one high-profile partnership, too. SparkCognition began integrating IBM’s Watson—the artificial intelligence computer technology known for competing on Jeopardy!—into its system last year. When SparkCognition recognizes a problem in something it monitors, a users can interact with Watson to research and troubleshoot the problem.
SparkCognition’s security focus is on preventing cyber attacks, in particular on physical assets. A hacker could infect a computer system—connected to anything from the motor of a vehicle to a nuclear centrifuge—with a virus that makes the device run slightly faster. Even if it’s just a few extra revolutions per minute, it could cause the machine to burn out, Husain says. The company’s software would monitor myriad data points about the machine, and be able to detect a slight change that a human wouldn’t notice, he says.
That’s the idea, at least. And it gets to the premise that Husain contends is the difference between his artificial intelligence software monitoring industrial equipment versus a human expert. The AI software is getting tons of data that the equipment pumps out, typically from sensors, and analyzing it in ways that a human brain wouldn’t—or couldn’t—because the sheer volume of data would shut it down.
The SparkCognition software is using deep levels of data to make comparisons and ask questions that humans wouldn’t consider because the depths of that data are so vast, Husain says. Artificial intelligence represents an evolution of computing, he says. We first learned how to collect data; we then learned how to communicate it and do so more quickly; we expanded the capabilities of storing data, and cut the costs of doing so; and most recently we have built a way to share it, with the cloud. Now, artificial intelligence will be able to harness it all, he says.
“The availability of data is so tremendous. The ability of the human being to be able to deal with cognitive burden is fixed and isn’t changing,” Husain says. “In order to make use of that data, the only implication is that all of the tools available in that category will go from being limited data, if-then-else tools, to cognitive systems that will not be told what to do, they will figure out what to do.”