NASA chief: AI will propel ultimate mission to search for life

NASA leaders met Wednesday for an Artificial Intelligence Town Hall to discuss how it plans to leverage new generative AI tools to align with the space agency’s overall mission. From data risks and replacing humans to searching for life on Mars, Cybernews has the story.

The live stream event kicked off at 1:00 p.m. EDT at the agencies headquarters in Washington, with opening remarks given by NASA head administrator Bill Nelson.

“We have used AI safely, sustainably, and successfully for decades. Yet today, this technology is transforming before our eyes,” Nelson began his address.

Nelson pointed out that when it comes to artificial intelligence, historically, the US National Aeronautics and Space Administration (NASA) has the most use cases out of any other government agency.

Nelson stressed that when used correctly, AI can accelerate the pace of discovery.

AI can “open new possibilities in our ability to land on celestial bodies, navigate to them, and peer into the vast corners of the cosmos, and even in the search for life.”

The search for life is a “statutory requirement” of NASA and why the agency is digging around Mars and searching for exoplanets, Nelson explained.

NASA AI town hall Bill Nelson
NASA Chief Administrator Bill Nelson delivers opening remarks at the NASA AI Town Hall live event. NASA headquarters. Washington, D.C., May 22, 2024.

The promise of new AI tools

NASA says AI tools are expected to help drive the agency’s mission by bolstering scientific research, analyzing data, and supporting spacecraft and aircraft.

“As this technology grows so does our capacity to use it, to test it, to refine it, and to integrate it into our work to try to benefit all of human kind,” Nelson said.

“There is new promise in AI,” said Nelson, citing new machine learning tools such as neural networks, deep learning, generative AI, and modeling.

The agency said it plans to integrate these new tools –as it always has – safely, security transparently, and responsibly, and in concert with the AI safety directives laid out by the Biden administration most recently last fall.

From a security system standpoint, it will also be business as usual for the agency.

‘We already have processes and procedures to securely protect the data we work with, now will we just be incorporate emerging AI technologies into our already existing secure NASA network system,” NASA officials said.

Deputy Administrator Pam Melroy also stressed the importance of responsible usage and the importance of robust governance and protection measures to safeguard against potential risks, such as unintended bias and data collection.

Melroy said the agency has already taken two concrete steps to minimize these types of risks, including appointing NASA’s first-ever Chief Artificial Intelligence Officer, David Salvagnini.

It’s second move will be to implement a government structure to develop the processes and protocols needed to handle, assess, and mitigate risks.

NASA teams have already been working to create these government structures, Melroy said.

Salvagni, who was appearing for the first time in his new role as part of Wednesday’s Town Hall panel, was hired by the agency last June as Chief data officer.

In addition to Salvagni, top brass on the Town Hall panel included NASA Chief Scientist Kate Calvin, NASA Chief Technologist A.C. Charania, and NASA Chief Information Officer Jeff Seaton.

AI still needs human oversight

The four experts fielded a wide variety of questions from staff audience members, from AI budgets to protecting the agency’s copyrighted images.

On the subject of deep fakes, Salvagni said NASA plans to “keep an eye on” the misuse of its images and where it posts content but noted that AI is pretty good at authenticating NASA imagery to prevent deepfakes from circulating.

The group spoke about the need to foster AI knowledge sharing and collective learning, not just within the agency but with the private sector.

Partnerships with the AI private sector, such as major scientific and research organizations will help drive innovation, NASA Chief Information Officer Jeff Seaton said.

“AI can inspire us,” the CIO said. “We don’t even know yet what insights we will get to see data in a new light. New AI tools will allow us to do things in ways we never imagined possible,” he said.

NASA AI town hall panel
NASA Town Hall panel (l to r) NASA Chief Scientist Kate Calvin, Chief Technologist A.C. Charania, Chief Artificial Intelligence Officer David Salvagnini. and Chief Information Officer Jeff Seaton. NASA headquarters. Washington, D.C., May 22, 2024.

Chomping at the bit to begin the process of integrating AI tools into NASA applications has been a source of frustration for the team, Seaton said.

Although AI capabilities are being built into many products, Seaton explained that NASA must wait for the version of these products that are approved by government agencies, which takes much longer than commercial products, such as Microsoft Co-Pilot, already being developed.

Once approved, NASA will have to examine which products will be adopted responsibility by understanding each AI system's capabilities, or lack thereof, as AI models have been known for fabrication, hallucination, and errors of omission.

Besides automating mundane tasks and summarizing large amounts of information, anomaly detection, such as wildfire detection and counting exoplanets, were some of the earth science applications mentioned among the panel that machine learning can accomplish faster than humans – a boon for innovation and advancing missions such as Artemis.

Observing the Moon and Mars with surface imagery features, communicating with spacecraft from large distances, and traversing rovers on surfaces further and faster will only become more accessible with generative AI, Seaton said.

We will have “a trove of information we can now learn from using mathematical models that can generate responses to our questions,” Seaton said.

Yet, the panel also made sure to address fears that AI would eventually replace human workers at the agency.

Even with the capability to ingest and analyze large amounts of data, human review of information gleaned from AI tools would be necessary to avoid mistakes by AI, they said.

Currently the agency is testing out its own large language model (LLM) and expects to have the environment ready to experiment with by mid to late summer.

The agency also announced it would be holding a new AI training summer camp for its staff to help them fully understand AI technology, data sets, and applying algorithms.

As for NASA’s AI budget, Seaton said funds would likely be determined by the value of each AI capability and how it would help successfully advance the 2040 workstream in a responsible way.

“AI will be leap after giant leap in the decades to come,” Nelson said.