‘Everyone is doing AI’: Space sector urged to catch up

COLORADO SPRINGS — The space industry must speed up adoption of artificial intelligence or risk falling behind, government and industry experts warned March 27.

During a symposium organized by the Universities Space Research Association and George Washington University’s Space Policy Institute, representatives of both the space and AI fields argued for greater use of AI technologies for both spacecraft and applications of space-based data.

Missions have incorporated AI technologies for decades, but early applications were rudimentary, experts said. Today’s large language models and machine learning systems offer far greater capabilities.

Steve Chien, senior research scientist and fellow at the Jet Propulsion Laboratory, cited software that allows space telescopes to skip or re-plan observations autonomously as one example.

“Twenty years ago, you would absolutely call that AI,” he said. “Now, you would say, maybe not.”

Chien was among those encouraging companies and government agencies to make greater use of AI in space systems. “The aerospace sector has not been as fast as it needs to be,” he said. “We need to change.”

Unwillingness to adopt AI technologies is increasingly an outlier.

“Everyone is doing AI. If you say that you’re not doing AI, people think you’re very strange,” said Rupak Biswas, director of exploration technology at NASA’s Ames Research Center. He cited the value of AI to analyze large data sets and extract insights that a human could not find.

Defense officials also see AI as essential to future programs, said Bryan Dorland, principal director for space technology in the office of the Assistant Secretary of Defense for Critical Technologies. They include using AI to improve space domain awareness, provide autonomy for missions in cislunar space where latency becomes an issue and supporting the proposed Golden Dome missile defense system.

“It’s gotten beyond the point where humans can be in the loop in processing information,” he said of those potential applications of AI.

Representatives of companies developing AI technologies also encouraged the space industry to increase its use of AI. “The capabilities are changing so rapidly it’s a completely green field,” said John Platt, a fellow at Google working on climate and science.

He explained how Google was applying its AI technologies at the intersection of space and climate, such as analyzing imagery of buildings to determine the optimal placement of solar panels. The company also plans to use AI to analyze data from FireSat, a constellation Google is developing with the Earth Fire Alliance and Muon Space to detect and track wildfires.

Simialrly, Meta is promoting its Llama open-source AI models for use in space. Laura McGorman, director of Meta’s “Data for Good” initiative, said an MIT project applied the company’s Llama model for a spacecraft AI navigation system, while another effort used Llama to process Earth observation data to count every tree worldwide.

She noted that Meta is not doing any specific outreach to the space industry on the use of its models, but acknowledged understanding what those models can do is an obstacle to wider adoption.

“You have this huge awareness problem at the very top of the funnel, and I think that’s probably the biggest barrier right now,” she said in an interview. “Expanding that level of awareness to space researchers is, I think at times, a tall order.”

Chien noted that traditional space companies are more reluctant to adopt AI than startups, who are more willing to experiment with the technology.

“They’re running ahead with AI,” he said of startups. “AI is now a branding.”

Platt encouraged an audience of primarily university researchers to experiment with AI.

“I don’t want you to be alarmed. I want you to be excited. No one really knows what we can do for any particular application. It’s so new,” he said.

“It’s just open. No one knows what they’re doing, and that’s great.”