Less than 18 months after telling Guardians to quit using ChatGPT and other emerging artificial intelligence tools while the service examined the risks and opportunities they posed, a Space Force leader said Feb. 26 the service has “done so much” to explore and expand AI adoption.
Seth Whitworth, acting deputy to the Deputy Chief of Space Operations for Cyber and Data described the progress during Booz Allen’s Space + AI Summit in Arlington, Va., as industry speakers touted AI’s potential benefits for domain awareness and command and control.
The Space Force ordered its AI pause in September 2023, alarmed that Guardians were using unproven generative AI tools without rigorous testing and proof that the tools were safe and reliable.
“There was a whole lot of unknowns, and we were championing ourselves as the innovation service,” Whitworth said. “There was fear that data would leak or we didn’t know those pieces, and so we said, ‘Take a strategic pause.’ We have done so much since that first memo went out.”
By June 2024, the Air Force Research Laboratory released its own generative AI chatbot, NIPRGPT, built for Airmen and Guardians to experiment with. Whitworth said his team hosted a “generative AI challenge” to identify more potential uses.
“We learned a whole lot along the way,” Whitworth said. “We were able to work with the Department of the Air Force and Department of Defense to re-establish some of those guidelines and ensure that we were moving forward in a secure way that didn’t hamper innovation.”
Guardians gravitated to the tool, seeing AI chatbots and assistants as helpful for everyday, non-operational tasks, like writing performance reviews and other reports, Whitworth said. Such back office functions have long been seen as ripe for automation, freeing operators to focus on sophisticated higher-level tasks.
But the volumes of data generated by sensors in space is such that automation is hard to ignore. Satellites generate imagery, signals intelligence, orbital data, and more, and with thousands of satellites and tens of thousands of bits of debris circling the Earth at 17,000 miles per hour, the need to collect, collate, and analyze data continues to grow exponentially, experts say.
“In the age of proliferation, especially with Guardians, one satellite to many [personnel] just doesn’t work,” said Nate Hamet, CEO of Quindar, a satellite operations company. “As we proliferate, how many people can we actually assign to dozens, hundreds of satellites, or where satellites are actually sending us information about what’s wrong and moving more of the anomaly prediction on board, at the edge, onto the spacecraft.”
Whitworth envisions a single operator aided by AI controlling multiple satellites. “I think back to Lt. Whitworth, who operated a satellite, and it was me to one satellite, and my partner next to me to one satellite. We operated that satellite individually. And that made sense at the time, because DOD had the largest constellation on orbit, and we were doing just fine. That very quickly exponentially changed as more commercial providers started launching more equipment, and the DOD itself pivoted to more resilient and proliferated architectures. No longer can I have one Guardian flying one satellite. There’s going to just be too many satellites, not enough Guardians.”
Indeed, the Space Force has already started automated satellite operations through its Space Rapid Capabilities Office, which is acquiring software for its Rapid and Resilient Command and Control system, and experimenting with it on test satellites in orbit, officials said in December 2024.
Whitworth cautioned that discussions and analysis are still ongoing about which satellite operations can be automated and what level of trust can be placed in the technology.
Building trust must happen gradually, said Dave Prakash, Booz Allen director of AI governance; if the Space Force can’t trust the technology, the whole process will break down, he added.
“It’s not about moving fast and just hoping nothing goes wrong,” Parkash said. “It’s not about being paralyzed … so that Guardians are now burdened with ‘not only do I have to use AI, but I have to do the manual process and double check it.’ [If that’s the case], I’m not actually making any labor savings. It’s actually this third option, where AI is a seat belt, not a speed bump, to accelerate delivering of AI for mission critical applications.”
Despite these hurdles, the challenges posed by growing complexity and data volumes make AI an attractive solution for mastering information from space, said Booz Allen executive Pat Biltgen.
“I think there’s a possibility that this domain could be enhanced by AI,” Biltgen said. “I think there’s promises that if you combine generative AI technologies that are grounded in physics, they can help solve this one Guardian, one satellite problem.”