With all the buzz surrounding artificial intelligence, officials at NASA say the technology is helping employees get their arms around all of their data, but the emerging technology isn’t going to put rocket scientists out of a job anytime soon.
Jeremy Yagle, the technical lead of the data science team within NASA’s Office of the Chief Information Officer, says after six years of experimenting with machine learning, the agency has leveraged the technology’s text analytics capabilities to make years of research more easily searchable.
“It’s not uncommon for somebody who’s worked at NASA for decades to have upwards of 10,000 or 20,000 papers on their personal computer. To be able to look at that really quickly takes time, and some of these new technologies reduce that time, and therefore save time for scientists to do what they really do best,” Yagle said May 24 at IBM’s Think Gov event in Washington.
While text analytics has freed NASA employees from hundreds of hours of low-skill work, the agency has much bigger plans for AI.
“Even the biggest journeys have to begin somewhere. We are the agency that went to the moon, we’re going back. We’re going to Mars. And so when it comes to investigating things, taking risks, failing forward, we’re certainly not adverse to that,” Yagle said.
NASA hopes to eventually develop a “Siri on steroids” that can answer complex math problems far beyond what someone could Google.
“Having systems that can answer questions for us in very complex domains doesn’t exist right now in very many areas. It’s something we really have our eye on,” Yagle said. “We see that coming, but we’re on a NASA timeframe. That means decades, not weeks or months or years. But it’s out there.”
In recent years, AI has become a major buzzword in the federal IT world, but most of the applications that exist today are low-level pilots that handle rote tasks, like copying and pasting information onto a form.
“As much as we try to look forward to what AI and cloud technology will do, we can barely see the next five years,” said John Kelly, the senior vice president of cognitive solutions and IBM research.
The push to develop more robust AI applications coincides with NASA’s agencywide goal of better data management.
“We need to build a knowledge base. It’s not enough to simply build a data repository or a database. We’ve got a lot of folks that are really good at collecting things, and nobody has any idea what was put in that file cabinet by the time it was done. We actually need to know that data is, how to pool that data out, how to connect it with other data sets,” Yagle said.
Last year, NASA Langley celebrated its 100th anniversary. The facility actually predates NASA —before 1958, the federal government called it the National Advisory Committee for Aeronautics (NACA).
Yagle says NASA has begun using machine learning to automatically detect cracks in advanced aerospace materials going through testing.
In the meantime, the agency has used text analytics tools to process and share more than half a million documents on space radiation studies among five research teams based in three separate NASA centers. NASA also shares this information with the Defense Department.
“What this has allowed that researcher to do is to see themes develop across domains, to know who important researchers are and to get all of those results in seconds, and to be able to visualize them in a meaningful way,” Yagle said.
Practical uses of AI like text analytics may not grab as many headlines as some of the more cutting-edge breakthroughs, but Yagle says it’s a tool that allows NASA employees to focus on more specialized work.
“The media like to talk about self-driving cars, facial recognition, predictive analytics … there’s a lot of things out there that are a little bit scary, a little bit hokey, a lot of buzz. This is one, when it comes to text, that we found a lot of value in down at NASA,” he said.