For Sale: Artificial Intelligence That Teaches Itself

SparkCognition’s CEO says his company’s AI might be able to diagnose and predict engineering problems like the ones that have bedeviled the USS Coronado and other U.S. Navy littoral combat ships.

US Navy

AA Font size + Print

SparkCognition’s CEO says his company’s AI might be able to diagnose and predict engineering problems like the ones that have bedeviled the USS Coronado and other U.S. Navy littoral combat ships.

An AI startup with dozens of aviation-industry customers sets its sights on the US military.

It’s not quite R2-D2 repairing Luke Skywalker’s X-Wing fighter in space battle, but artificial intelligence could soon be helping the military predict when equipment will break, fend off cyber attacks, and prevent ships from colliding with one another.

That’s the promise of AI systems by SparkCognition, an Austin, Texas-based startup. Founder and CEO Amir Husain says his AIs can teach themselves enough about a field of endeavor to diagnose situations and offer solutions. Or, as Husain put it: “Our algorithms can extract the physics of the problem just by observing the data.”

SparkCognition, which is already providing services to dozens of aviation-related firms, recently received investments from Boeing and Verizon as part of its initial $32 million funding round. The company has also attracted the interest of former and current Pentagon officials. Retired Marine Corps Gen. John Allen is a board member. Among the firm’s senior corporate advisers is Wendy Anderson, who served as chief of staff for Deputy Defense Secretary Ash Carter and as deputy chief of staff for Defense Secretary Chuck Hagel.

The Pentagon has made no secret that it wants to pair humans with machines to help them make decisions faster. The military has been looking for ways to automate intelligence processing using a new algorithmic warfare cell. The work done by SparkCognition appears to fit into the sweet spot. Already, these algorithms are being used by commercial firms to forecast failure rates for ship turbines and pumps. And they can predict more than when equipment will fail, but also the type of failure and why it’s failing.

One customer is Flowserve, a Texas-based pump manufacturer.

“We built a solution for them that extended their failure forewarning from four hours to five days,” says CEO Husain. “That’s the kind of impact [the company’s algorithms have]. We don’t know anything about pumps; we’re not domain experts in pumps.”

In the aviation industry, he says, SparkCognition is supplying “a huge array of now close to 60 of the largest industrial aviation customers…All of them have pointed to the uniqueness of this automated model building approach.”

The group includes Honeywell and new investor Boeing. “SparkCognition is at the forefront of a technological shift in machine learning and artificial intelligence that will revolutionize every aspect of industry,” said a statement by Greg Hyslop, Boeing’s chief technology officer and its senior vice president for engineering, test, and technology. “They are leaders in AI, and they are pursuing the types of technologies that are critical to our future products and services.”

In the defense sphere, the company has used its “natural language algorithm” to transform military manuals for a defense firm.

“So instead of going up and looking through these manuals and figuring out what to do when something breaks and going through that prognostics process, you want the model to basically anticipate what you need and also fetch information that is collected from various different manuals … based on the intelligent understanding of that content and is provided back to the person asking the question,” Husain said.

From a military perspective, the algorithms have much potential, particularly in cyberdefense.

“The threat surface is so massive that dealing with the types of threats that are emanating and targeting the cyber capabilities of any country of any developed economy, that is your one area where the use artificial intelligence to block that is a huge contribution,” Husain said. “We are working on that.”

Husain said his algorithms might also be able to help with the reported pump and turbine problems with the Navy’s Littoral Combat Ships. “That is the exact kind of thing we look into and that we protect on a commercial level with a large number of companies today.”

Then there’s the R2-D2 applicability across all military systems.

An “AI watchman” could prevent ships from colliding with one another since the computers are “constantly looking at sensor data and is making sense of the environment and the situation.”

The company is now “investing heavily” on a project Husain calls automating decision making, using artificial intelligence to plan and take action.

“There is that safety aspect of using artificial intelligence to augment the level of capability and intelligence available on ships, on tanks, in aircraft, all over, where you almost have an embedded AI technician be part of every military asset,” Husain said. “That is a capability and it leads to benefits that are tremendous. It leads to not just monetary savings, but the savings of lives.”

And the possibilities may be endless. Husain says there’s an easy answer to the question: Where can AI be applied? “It can be applied literally everywhere,” he said.

Close [ x ] More from DefenseOne