The National Geospatial-Intelligence Agency harnesses the power of artificial intelligence and machine learning to address the challenge of acquiring, processing and managing large amounts of data in a short period of time.
NGA notably uses AI and machine learning to detect and decipher satellite images while actively developing new ways to manage the data made available by these tools. In this mission, the people using AI and machine learning are equally important in the process when it comes to addressing challenges related to bias in collected data, managing result drift, and maintaining security while advancing technology.
“I think one of the big goals, as well as the exciting opportunities of AI, is to help people do their jobs better and faster, and even on a daily basis, like supporting people who code and using great language models to help them create code. . It’s not that these models are going to replace us in writing computer code, but they can be a tool,” said Natasha Krell, a specialist in computer vision, machine learning, research and development at the National Geospatial-Intelligence Agency, on Federal Insights – Artificial Intelligence and Machine Learning. “I think someone said recently, it’s almost like Word document processing. The Word document will not write the text for you. You still have to type it, but it’s a tool that helps you complete your work.
NGA leverages these emerging tools to find that proverbial “needle in the haystack” when analyzing images.
“It’s really useful for detecting these objects and then classifying what they may be and, given the deluge or the large amounts of data coming from satellites, AI is really a tool to process all of these images and, again once, the information that comes from it,” Krell said. Federal Drive with Tom Temin. “We are exploring on-premises and cloud opportunities across storage, infrastructure and compute. »
NGA classifies its large multimodal models for imaging into several different types, including electro-optical, synthetic aperture radar (SAR), and thermal. Bringing all of these together is what Krell called “the next step, and really the cutting edge of what’s happening in the field of AI.”
Krell said NGA and its partners are using AI and ML to impact the humanitarian aid and disaster relief mission space.
“We have some AI and machine learning initiatives that we can share publicly and work with other government partners in different organizations,” she said. “We certainly share models of results, data and information. There is certainly a good track record in this area.
Like most agencies, NGA faces its share of challenges integrating AI and ML.
She said staff training, managing the volume and velocity of data, and protecting against bias are just some of the key concerns.
“AI, in general, is very data intensive, because the data management systems and processes are really important. There are also interesting developments and I would even say paradigm shifts with these foundation models. These are large pre-trained models that are the starting point for natural language processing (NPL), and you essentially adjust it based on a specific use case. . .the point of view of taking into account a large model that is already pre-trained and then fine-tuning it on a smaller dataset,” Krell said. “Bias can come from many places, whether it comes from the data sets, the models, and then the algorithms themselves. So it’s definitely something to be aware of and something we ultimately have to balance.
Additionally, NGA should be aware of the security risks associated with these tools.
“You can’t just take any machine learning or algorithm off the shelf. You really need to do rigorous verification and validation of the models and the data that you bring in,” she said.
NGA relies on advice and documentation from the Chief Data and AI Officer (CDAO), the Air Ministry’s Main Data and AI Office as well as the Joint Artificial Intelligence Center to ensure that AI and ML tools are used appropriately.
Krell said that while NGA is well aware of the limitations and challenges of AI/ML, the growing use of tools like ChatGPT has helped the general public and its employees better understand the technology.
“There’s a lot of interest in natural language processing in academia and industry…And there’s some really exciting progress happening in industry and academia, particularly in the area of large multimodal models. So this is where you use both the text and the images to maybe query the images and get more information about the results that the machine learning is generating,” she said.
Copyright © 2023 Federal News Network. All rights reserved. This website is not intended for users located in the European Economic Area.
