AI chatbots like ChatGPT continue to attract attention for their ability to churn out essays, reports and emails. For space companies, generative artificial intelligence tools offer the potential to merge datasets and streamline operations.
At HawkEye 360, for example, analysts who query the company’s virtual warehouse of radio-frequency signals through programming code could instead find what they want by asking natural-language questions.
“Show me the illegal fishing that’s happening in this area of the world,” Kaitlin Zimmerman, HawkEye 360 chief data scientist, said in October at Satellite Innovation 2023 in Mountain View, California. “That could generate the query that could reach into our database and pull that information out.”
Similarly, hyperspectral startup Orbital Sidekick allows employees to search the company’s imagery archive and tasking schedule using generative AI.
“It’s a really powerful tool for us to watch how our teams interact with these systems,” said Andrew Guenther, Orbital Sidekick principal software engineer.
In addition to seeing what information people want, Guenther and his colleagues see when queries come up empty. “They’re writing down exactly what they want to do, and we can capture that,” Guenther said at Satellite Innovation.
Transforming work
For years, space companies have been expanding their reliance on machine learning algorithms to automate operations, reduce latency in data transfer and detect anomalies onboard satellites. Those applications of traditional AI can be designed in advance and tested thoroughly.
In contrast, generative AI relies on deep learning models to detect patterns in enormous language or imagery datasets and generates results based on historical data and future predictions.
“Obviously, generative AI has taken the world by storm,” Steven Truitt, Microsoft Azure Space principal program manager, said in October at the MilSat Symposium in Mountain View. “It’s going to have huge effects on the way that the space industry as a whole works.”
For instance, it could lead to better integration of disparate datasets through machine-to-machine communications.
“Even ill-defined APIs will be able to talk to each other by virtue of a generative AI translating service,” Guy de Carufel, CEO of Cognitive Space, a startup specializing in automating satellite operations, said at Satellite Innovation.
In addition to transforming the way space companies work, generative AI will have a profound impact on relationships with customers.
“It’s going to fundamentally change the way their organizations work, the way that their patterns of behavior are manifested in the world, and that’s going to have carryover effects,” Truitt said.
Solving complex problems
Customers of Esri’s ArcGIS web-based mapping software, for instance, can become overwhelmed by the amount of data produced by Earth-observation sensors.
“When you look at all the sensors we’re launching and all the data that’s coming down, there’s literally no way we can consume it all,” Richard Cooke, Esri global business development director, said in September at the World Satellite Business Week conference in Paris. “The work that Synthetaic and other organizations are doing will set the stage for that next application of generative AI. We will get to the point where all these sensors are going to operate like a central nervous system for the planet.”
Once the sensor data is linked, “it’s going to be like neurons firing in the brain,” Cooke said. “We’re going to get observations that quickly, and we’re going to be able to make decisions on those observations that quickly. It’s going to enrich policymaking [and] business decisions and improve productivity.”
Synthetaic is the Wisconsin-based AI startup that traced the path of the Chinese weather balloon shot down by the U.S. Air Force in February by merging social media reports with Planet satellite imagery.
With generative AI “is going to come fairly radical transparency,” Corey Jaskolski, Synthetaic founder and CEO, said at World Satellite Business Week. “We’ll be able to find things in satellite data that in the past were hidden in obscurity.”
Radical transparency promised by generative AI could help people solve complex problems.
After a hurricane, earthquake or other disaster, generative AI could synthesize satellite and terrestrial data to determine where communications networks and transportation infrastructure have been damaged, what resources are available and recommend courses of action for emergency managers.
Installing guardrails
What’s more, generative AI can test future scenarios.
Amazon Web Services is working with Latitudo 40, an Italian Earth analytics company, to help urban planners model new towns.
“They can reduce their carbon footprint to a minimum using generative AI with Earth-observation data,” Alan Campbell, AWS principal space products solutions architect, said at Satellite Innovation.
People exploring potential applications for generative AI also warn of potential problems.
Companies that employ large language models to simplify database queries will need guardrails to ensure customers can only access data they have permission to view.
“We’ve been taking a hard look at our security designs to separate those datasets and provide hard boundaries to make sure we don’t expose any information that we shouldn’t,” Zimmerman said.
In Earth imagery, generative AI models could fill in gaps with fabricated data, a problem known as hallucination. Or data could be intentionally manipulated to produce deep fakes.
“AI can help us be so useful in the security space,” Sue Gordon, former principal deputy director of national intelligence, said at the Economist Space Summit in Los Angeles in October. “There are others who would be using it to try and subjugate our systems to bad purpose. We need to really double down on cybersecurity, including data assurance.”
The message was reiterated in an Oct. 30 Biden administration executive order. The Executive Order on the Safe, Secure and Trustworthy Development and Use of Artificial Intelligence warns that AI poses security risks, including making critical infrastructure systems more vulnerable to physical and cyber attacks. And it directs the Departments of Defense and Homeland Security to explore how AI could enhance cybersecurity defenses.
Predicting interference
Military experts see enormous potential in both traditional and generative AI.
Retired U.S. Air Force major general Kim Crider said AI can augment electronic warfare defenses by consuming massive datasets and detecting interference patterns.
“Machines help us understand where are these interferences coming from or where will they come from,” Crider, founding partner of consulting firm Elara Nova, said at the MilSat Symposium. “We can predict the nature of the interference. We can predict where the interference will be. Certainly, there are risks, and there are going to be challenges. But we need to figure out ways to continue to advance AI and quantum and a variety of other capabilities that will move us to the next level.”
This article first appeared in the November 2023 issue of SpaceNews magazine.
