Artificial intelligence arms race accelerating in space
WASHINGTON — One of the next big things in geospatial intelligence is tiny black boxes aboard satellites that ingest massive amounts of data in space and instantly analyze it. No downloading necessary.
Geospatial data manipulation and analysis in real time is the holy grail in the military intelligence business. “We are trying to help commanders ‘see through the fog of data’ in situations when they have to make decisions very quickly,” said Melanie Stricklan, chief technology officer and co-founder of Slingshot Aerospace, in Manhattan Beach, California.
Stricklan served 21 years in the U.S. Air Force, and her duties included flying in the back of the JSTARS radar surveillance plane. The airplane’s sensors were pulling loads of data but it was hard to extract intelligence, she told SpaceNews. There were times when the data would be sent to analysts and it could take weeks to actually “see what we were looking at.”
Later in her career, Stricklan worked at the Air Force Space and Missile Systems Center on satellite programs. It became clear to her and other colleagues in the industry that the future would be won by those who could figure out how to use technology to exploit all that data collected by satellites and airplanes.
Slingshot Aerospace is developing a cloud-based platform that pulls in data from many types of sensors and uses machine learning algorithms to “extract information,” she said. “Our end goal is to have AI in a small chip on board spacecraft.”
The technology is still in its infancy, to be sure. The company continues to bring more data into the cloud-based platform to try to improve the algorithms. The goal is to put the platform in a chip that would be embedded in satellites, said Stricklan. “We’re trying to develop this so the extraction takes place on board the spacecraft.”
The military for decades has been interested in “data fusion.” The phrase is taking on new significance as the remote sensing industry moves to deploy clusters of satellites that collect multiple forms of data and signals, from traditional high-resolution imagery, to radar, radio-frequency an hyper spectral pictures. “The only way to get through that amount of data and make it relevant in the decision space that a warfighter needs is by using machine learning and AI,” Stricklan said.
The geospatial intelligence business used to be about selling data. “Now everybody wants solutions,” said Robert Laudati, managing director of commercial products at Harris Space and Intelligence Systems
A confluence of trends is changing the industry. Data is becoming easier and cheaper to collect. AI has been out there for years, but now the industry is beginning to “operationalize it,” Laudati said in an interview. The military and the intelligence community are fueling the demand for more advanced algorithms that can take advantage of any data, regardless of the source. “How do we stand up the systems to drive this massive amount of data and do something about it?” That is the question of the day.
With so much data being gathered by remote sensing satellites, the “real future” is to move the computing to space, Laudati said. Experiments already are being done aboard the International Space Station. “If you do the computing up there, you don’t have to download so much data,” he said. The problem is that satellites were not designed to be computers. He does not see that capability coming to the market any time soon. “That’s a long way off.”
The urge to operationalize AI is one reason the Pentagon has committed to invest $10 billion to modernize its cloud computing platforms. This is crucial to algorithm developers and geospatial analytical companies, said Stricklan. “DoD has many cloud challenges,” she said. The military needs cloud environments with layered security levels so vendors can be “play inside” and understand what customers needs. “That’s a barrier to entry for startups,” Strickan said. “We overcome that by being cloud provider agnostic.” Regardless, it is “frustrating,” she said. “DoD is far behind in understanding how to use the cloud at scale and how to do it in a multilevel security environment.”
Deputy Defense Secretary Patrick Shanahan said the Pentagon’s cloud procurement is all about “how we deploy artificial intelligence and evolve its capability.” The first step is to “have an environment where the data is hosted,” he told reporters last month. “So I would characterize the cloud as ‘we’re ushering in a new age of technology.’”
During a recent Senate Armed Services Committee hearing, Sen. Martin Heinrich (D-N.M.) pressed Defense Secretary Jim Mattis to explain the hurry to award a cloud contract by September. “The department seems to be rushing ahead to issue an RFP in early May and intends to issue an award as early as September,” Heinrich said. “What is the rush here? And why is the Pentagon moving forward so quickly, despite the concerns of both Congress and technology leaders?”
Mattis explained that DoD regards the cloud as a vital weapon that commanders need in the battlefield. “Senator, the rush is that we have too many databanks that the front-line commanders cannot swiftly draw information from,” he said. “What we have been looking at right now is how do we get faster access for the young folks on the front lines and displaying the information they need — not all the information in the world?”