Closing Keynote: NASA’s Vision for AI in Space Exploration
So uh good afternoon, everyone. Thanks for um staying um uh at the summit. So to listen to me talk um before I start, I um I did want to say something probably uh as an assign. Today is uh at NASA, we are celebrating the day of remembrance. So, every year
and we take the whole week at NASA to think about um the previous journeys and experiences we've had at NASA and the lessons we can learn in terms of safety and a safety culture. Um So we remember uh Columbia Challenger and Apollo one. So we actually take this week to have um our employees um uh kind of listen to our leadership reinstall the culture of safety, but also innovation at the agency. So,
uh uh today is a special day we've had um uh this week we've got different ceremonies um at Arlington National Cemetery Cemetery. Um And with our workforce. So uh that this is a uh uh a week where sometimes at NASA, we reflect on the past now, today, we're actually looking at the future. And so, uh as I was telling um someone earlier, this is a good book. And for
you all, I think if for those of you that stay today at the beginning of the day, uh We look back in terms of the James Webb Space Telescope, the incredible achievement, how we achieve that, the data that came out of that. Now as we look forward, how do we drive that innovation um within the agency and within the nation? So first I'll start off a little bit about me. I'm not gonna powerpoint slide you to death here today. Um I'm the agency, chief technologist or AC T as I call it uh at NASA headquarters. Uh
We support our senior NASA leadership uh Bill Nelson Pam Maroi and Jim Free in terms of technology strategy, uh policy uh and economics, in terms of our particular office at the agency uh at NASA headquarters, I'm in the Office of Science Tech uh Office of Technology Policy and Strategy uh within the A suite. Uh And basically look at um uh several moves ahead in terms of the agency in terms of technology disruption, uh places we can infuse technologies and how we're investing within the agency. And I've been at NASA about a year, first year as a civil servant uh coming from commercial industry uh 23 years in uh space and aviation. So prior to joining NASA, I
was actually at an aviation autonomy start up called reliable robotics where we were trying to automate the world's cargo airplanes to fly themselves. Uh It's a very interesting problem there uh dealing with human machine interfaces. Um prior to that working for Jeff Bezos at blue origin uh in terms of maturing the lunar lander program at blue. Um in that sense, trying to have a large cargo Landers and crude Landers that uh use machine learning and other algorithms to land humans on, on the moon. So, uh prior to that worked for Richard Branson for five years at Virgin Galactic.
Uh And that's particular particular instance, looking at the launcher one program, uh basically looking at autonomous rock and the left wing of the 747 and launching small satellites to space. So, um and prior to that 12 years at uh a uh a company in Atlanta called Space Works, doing a lot of advanced concept systems analysis, starting new companies. Um And actually part of that um started at Georgia Tech in Atlanta, Georgia. Uh And actually Doctor Bilen right
there who wrote a new book, I believe on A I uh that I think you'll find here. Actually, he and I went to school together. So it's good, good to see Doctor Bin here um at, at Booz Allen. Um So I
kind of feel a little bit at home uh with, with Pat here. So um so at NASA as agency chief psychologist, my portfolio is to look at our technology strategy, look at where uh we can leverage the disruption that's happening. So things coming into the agency actually two major technology areas um I've been focused on beyond some of the others that you might traditionally think of in terms of technology, those include quantum sensing and A I. Those have actually over the last year, our of us in particular has been supporting the agency thinking about those technologies more strategically as a chief technologist. Though I'm not uh uh kind of responsible for cybersecurity or it, we do have a Chief information Officer, chief data officer at NASA. And I work very closely with those uh
entities and personnel and staff and leaders on those kinds of collaborative areas. So for instance, an area where we collaborate together is an A I, we look at A I obviously for our uh it functions and, and other things, but also A I for engineering, digital engineering, machine learning and how we use it. Uh And that also brings together our chief scientists. So actually our Chief sciences at NASA, our CIO myself and others, we many times collaborate together of how we as an agency can leverage these tools and technologies. All right. So let's see if this works and slide. All right.
So um once again, I'm not gonna PowerPoint you to death. Yeah. For, for me, as I was thinking about this, how are we using A I? And really in the day for us it's our strategic objectives. And ultimately, uh if you look at the NASA vision, it is, it is to help humanity and benefit humanity. Now
we decompose that objective into looking at uh Science National Posture and inspiration. Both that is the knowledge of we as humans have. How do we help us industry, academia and how do we inspire those three pillars of Science National Posture and inspiration kind of drive a lot of our thinking of what, what we're doing and there's decomposition of those objectives and stakeholders that play into that. So for instance, in science, we have a National Academy, the cl survey process where we uh canvass our scientific community for the top scientific objectives we should attempt in planetary science and helios its uh and earth science. And so there are processes that drive the scientific objectives and those that get vis uh promulgated into actual missions or science missions, space science missions, planetary missions, human space exploration. We also uh mature technologies. So different parts of
the agency are working on research development internally with NASA civil servants as well as with industry and academia. Now, in all those three areas, we've been using A I machine learning for years. So in terms of generator A A I Yes, you know, last year, I think we were all there when ce PT came on board, I was at reliable robotics and it was an awesome time to be in the room when the coding, coding team realized this was available to them. It was like a eureka moment to still recall that moment very, very crisply last year when I was in industry. But separate from that, we've been using A I machine learning for many years at NASA from Earth Science data um to how we're looking at different data sets and missions.
So for us, it's not a new thing specifically our, our science portion of NASA. But now we're looking at uh digital engineering, Digital Twins, as I mentioned in the previous um uh panel, we've been looking at that for many years. And how do we use that in missions? And now how do we use these large language models obviously? But also how do we couple A I with, with things like 3D manufacturing. So different examples of how we're using A I or been thinking about using A I as we go forward are in these actual missions. So as as we look in autonomous missions on other planets on the lunar surface or on Mars, we've got a helicopter on Mars. That's the ingenuity
helicopter. You see that's flood more than 60 missions. How do we in future generations of Mars helicopters that we're looking at instill more autonomous operations to decouple such uh helicopters from, let's say the rovers that, that they are uh leveraging in terms of um like a base station. So those kinds of missions we're looking at, we've got missions on the lunar surface, looking at autonomy. We've got a program with Gapl called cadre that's looking at a collaborative set of rovers that use autonomy. And also as we look out further in terms of decreasing the life cycle cost of these missions, how can autonomy A I support that kind, those kinds of missions and in operations, we also look at science obviously and a lot of that science is looking at earth as observational data in many different wavelengths uh from earth to uh the planets of our solar system. This particular picture is of
a project we did with Nasa God or and academia that looked at uh losing a human train data set to then identify and characterize how many trees are in Western Africa. So a human train data set was used on tens of thousands to identify tens of thousands of trees. Then that data set was used to identify more than a billion trees and just the population of those trees are potentially arid areas. Now, that data can now be used to help determine the carbon footprint of all those trees in Western Africa. Um And we are using those kinds of uh uh A I machine learning techniques all across the board. The other thing I uh we think about both in terms of myself, the cio and uh other parts of their agencies as we go forward in time.
How do we work more digitally with all these tools? So machine learning, uh Digital twins model based system engineering A I, all those tools uh coupled with um kind of physical tools like we have with 3D manufacturing uh to actually work faster. For me the 21st century in aerospace and in the agencies also, how do we execute rapidly bring decision making down to lower levels and and can execute rapidly on these missions with the constrained budgets that we have and working digitally, I think will allow us to operate faster, reduce overall life cycle cost and hopefully um develop safer systems. So this actual picture um uh I think is representative of projects we've got ongoing at NASA Goddard where researchers are coupling A I machine learning with actually 3D printing to then 3d print uh structures that humans would never have imagined that would be more rigid or way less.
And so thus, the A I machine learning capabilities are developing things that will make these systems lighter and cheaper that we would never have imagined as humans to create. So I think a lot of tools and techniques that I think beyond just large language models and generative A I that we can provide to our civil servant workforce and to the industry and to, and to then promulgate that out to the contractors that are supporting us at NAS A next slide. And then here I've got different examples and I'll probably go through of different uses of A I and ML. Some
of these are, I would say um ex examples that we're using some of these, we've actually used an actual mission. So these kinds of examples range from uh kind of the observational side. So starting from the right hand side, we've been using um this particular example is kind of cool. And uh Doctor Viler won't like this. This was actually a collaborative project between NASA Langney commercial industry and Georgia tech uh where we actually took uh some data from uh Calypso Liar. Calypso is one of our spacecraft um that you see there, we took some of the data that was coming from Calypso and uh work with academia and researchers to look at how could we identify the health of coral reefs. So for instance,
a lot of researchers are looking at which coral reefs on the planet are at most risk. And so this, in this particular case, we were using Calypso data not particularly developed to help in this problem with uh trained data sets to identify. OK, which of the coral reefs that the Calypso data is observing are at most risk. And so that was kind of a cool experiment we did. Um And uh I think the results of this were published of this particular show were published in 2021 during COVID.
But it was kind of a cool test case study of using earth observational data in partnership with academia to look at a problem. The original instruments were never designed uh or optimized uh to, to solve. So which is kind of cool, there's multiple other examples of using Earth observation data in this way, in terms of um A I and ML another example I kind of like is the middle example. So uh OK, landing on Mars is pretty difficult and a lot of nations have tried uh the last few times. Uh we've been pretty successful and I think we're celebrating 20 plus years um since the Mars Pathfinder landing. So kind of a cool anniversary, but
it's pretty hard to land on Mars, even harder to land in terms of a particular place you want to land in. So if you're trying to land in a place and return samples back to earth, you want to land in a place where you think you'll have pretty good samples that might have evidence of possibly life on the surface of Mars or, or, or uh historical life in that sense. So you need high landing accuracy. Well, where are the
ways you can do? That is actually a way we've been using and we've used on the perseverance rover to land it on the surface of Mars to reduce what we call the landing ellipse. So a landing ellipse is the uncertainty of where you might land uh on the Martian surface or the lunar surface. And technologies where a camera system on this vehicle that's entering the Mars atmosphere and coming down is comparing the pictures it's seeing with the data sets we've gathered from previous Mars orbiters and during that comparison, it can then tell itself where it is. There's no gps on MARS. So you got to figure out where you
are. And so using these terrain, relative navigation technologies for identification of landmarks to seeing where you are maneuvering yourself to land and reduce the landing ellipse error that you have on the Martian surface. So we're using similar techniques, we're thinking about using similar techniques uh like trn on the lunar surface. We are
funding several. Uh We are the anchor tenants, anchor customers of several small lunar Landers that are moving uh gonna land on the lunar surface and they're using some of these technologies as well. Another reason why I'm kind of excited about that one is when I was at blue origin, we actually tested some of those technologies on New Shepherd, which is the vertical takeoff, vertical landing sub orbital vehicle at blue origin.
So once again, we've done terrestrial tests with Blue origin Masten on these technologies. And I've actually flown these technologies on Mars and looking to apply these technologies and comparing orbital data sets uh with the real time images these Landers are seeing um to help them land more accurately. So kind of cool there. Uh Another one is a period table of life uh project called Pedal. And this is kind of a open source framework uh that was developed a few years ago to look at how to use A I to help examine linkages in large language models relative to biological data. So once again, a lot of stuff uh people are looking at, but this was actually even a few years ago. So before chat GP T, we were actually developing these kinds of frameworks um to look at how to um uh how to help using biological models, understand some of these large language models.
So kind of a cool experiment there. Once again, the idea of using a IML for earth uh is a great application, but we also explore the rest of the solar system. So in this case, we're also looking at how we can use a IML to look at other planetary bodies. Uh What excites
me about this century is the achievements that NASA could possibly help it. And some of those achievements include our human exploration goals uh of going to the moon and landing humans on Mars, but also extends to future observatory. So you have um uh we talked about James Webb. I also hope that in this century we can also image an exoplanet. And so we have a habitable worlds observer program whose objective is that basically to return a picture uh to humanity of a planet in another solar system. But in addition,
there are other um very cool uh missions regarding uh life that could be here in our solar system. So Europa and Enceladus, the moons of Jupiter and Saturn are icy worlds where we think there might be hydrothermal vents underneath ice layers uh that could be areas of possible life. So, in this particular case, can we use images of Europa uh one of the moons in our solar system to detect different ice plates and use A I AND ML to help detect those ice plates. It's kind of cool application. We've
also got missions going to Titan and we are interested in the composition of Titan. And so using A IML to take uh observational imagery from Titan to observe methane cloud detections or formations. Another example of how we're using uh possibly a IML to help in the search for life is JP L uh has looked at how to use A I and ML to help uh develop autonomous underwater robots to be used on Europa. So once again, we're using a IML to help us identify ice plates in Europa. We're also using
a IML in missions that we're formulating maybe two or three decades from now to look at autonomous systems that would then go through the ice um uh melt and go explore these ocean worlds. These icy ocean worlds was actually kind of cool about that. One is when I was a freshman at Georgia Tech uh with Doctor Bilen, that was one of my first undergrad your design projects, which is to look at a Europa Lander mission. So it's kind of cool. We never, I think I imagine 25 years ago we'll be using a IML to help us develop the missions, analyze the orb imagery. And then also use that to help NAV help these robotic underwater submarines navigate these ocean worlds. So kind of a cool set of examples of how we're using A IML across the agency. Um You know, and uh as
we, if you want to know more, there's obviously uh you, you can do a lot of searching uh all federal government agencies are, are many of them also put out an A I inventory. So if you actually search NASA A I inventory, you can get a list of some of these projects that we're working on relative to A IML. And that's an annual inventory. We update other other federal agencies update that as well. Another resource for you
all to understand a IML autonomy technologies we're developing at NASA is a website we call Tech port.nasa.gov uh where we list many of the technologies we're investing in. So if you're kind of wondering what are you investing in at NASA? Well, Tech port.nasa.gov is a
good website for you all to go to, to understand not just a IML but a whole whole range of technologies. Um And the other thing, I think we all uh relative to this summit and the government we're wrestling with is a IML and our strategic understanding, strategic use of that. And so, for instance, I think uh earlier today, there were conversations about the executive order and uh an OMB uh letter relative to all federal government agencies. Um And so I think if I came back here a year from now, uh all those uh White House stakeholder interests for A I and ML will result in a chief A I officer at many of these agencies.
So actually, I would encourage you next year, probably to have a, a chief A I officer across the federal government panel because I think we will have named many of these agencies, those those individuals and uh and uh define those roles and responsibilities, which is kind of cool to think about. So, um not saying what panels you should have next year, but uh that might be a good one to have because we'll have those named individuals. Um It's a pretty exciting time at NSA Relative to the challenges we have of human exploration science, urban air mobility, which we also work on and leverage some of these autonomy, technologies and A I technologies. Um And going back to the moon uh for me, um you know, and I was speaking to Pat earlier, you know, it was really uh strange to think that, you know, 25 years ago, we were at Georgia Tech and who knew that? You know, I'd be at NASA while we're headed back to the moon and, you know, Pat's writing a book on A I, you know, uh couldn't imagine and stuff like that, right? Um But for me, it's all about uh kind of not just the inspiration science, but the strategic objectives of these kinds of endeavors. And what I'll leave you with is all of these technologies that are helping humanity. Uh And particularly for me, I look
at the lunar surface and lunar exploration, which is pretty important at NASA and to our stakeholders and leave you with the final thing of what A IML and all these sorts of technologies can get us to, which is what I would call the lunar moment. Uh As we look at investing these technologies and lunar exploration. For me, it's also about the moment that uh humans walk on the lunar surface. And from that moment forward, there will always be a human on the lunar surface, right? And these kinds of technologies will enable us to enable those humans to more permanently for longer term uh and sustainably live in these planetary bodies and live in space. I think it's a pretty powerful tool that I think 25 years ago we would never have imagined would be in our tool set. I think the machines we designed uh the data we look at uh the experiments we do. Uh and the knowledge we gain will be
enabled by all these tools and we're a part of it. So uh thank you very much. All right, maybe one or two. Yeah. Yeah. So, oh Pat, yes, Doctor Belcher, please. I got. So I see. I
do have a question. So you know, you were doing commercial space like before. It was cool for, for school. Um I've really been working
that for a huge chunk of your career. A lot of start ups, innovative concepts, high risk things. Now, I have read that sometimes when people with that background going to the government, there are cultural challenges because the government has structured processes, ways of doing things. So when you were appointed chief technologist, I thought like this is really exciting both for you and for NASA. But can you give us, I'm just really curious on your perspective of how you bring that commercial perspective into NASA and then either challenges or approaches to try and help a federal agency adopt a commercial mindset. I, I think once again, I'm not coming into NASA as a uh novice, I've dealt with NASA for 23 years, 25 years. Um And NASA part of
NAS A should pay for grad school when I was at Georgia Tech NASA funded our research lab. And so I'm kind of a product of, of the space agency in some sense. Um So I think I have, I have submitted contracts ranging from $50,000 to 6 billion to the government. You know, I propose that I've been involved in lobbying stakeholders, I've been involved in putting technical teams together and understand that ecosystem of, of working programs with stakeholders, with the government uh with NASA. And so, and, and going to NASA headquarters in my career and other centers. So I was pretty familiar with the ecosystem and how NASA operates. So that didn't
really surprise me. So I knew what I was getting into. So, first of all, uh because people warned me or there, you know, it's like, oh, are you sure about this? So, um so I think having that experience is helpful, um I also probably an approach I use is probably uh um uh uh uh kind of a uh uh platonic kind of approach of questioning. So once again, you know, these meetings, you kind of uh in order to get people to maybe reach a conclusion, you ask, well, from fundamental principles or first principles, why this, why this, why this? Um And then I, and so I think I can also point to successes.
So the CS program with commercial cargo to the International Space Station, commercial uh crew resupply uh uh to the International Space Station, these other models that have worked successfully, those successes help us advocate for additional public private partnerships. When I was in industry, we helped create public private partnerships from um I would say from scratch, but in collaboration with the government and so those successful partnerships, uh we can now point to, to continue the momentum of these partnerships. So one of the reasons I came to the government was to continue this momentum of partnerships and say, hey, these have been successful. Let's start expanding them. And so now what you see the human landing system program, which is a public private partnership of spacex and Blue Org and developing propellant depot based single stage lunar Landers for humans. Those are not architectures that NASA probably would have come out with on its own. But in co collaboration with industry and
their cof funding, uh we can now work with them to mature those technologies support them where we can. And so I think those smaller successes over the last two decades, conti and continuing those successes in low Earth orbit and beyond uh using first principles, I think. Um you know, uh and also I, I think finally, for me, Dyson on the cake of why to take this job um in the government is also leadership. So um I don't know if many of, you know, our deputy NASA administrator, Pam Melroy, you know, that was, I knew Pam when she was at DARPA and that was probably the icing on the cake for me to take this job of. Uh there's a spirit of
public private partnership at NASA. I know that because I was part of that in industry. Um you know, we have great partnerships with companies that didn't exist 20 years ago that are pretty substantial and have great capabilities. And finally, the internal leadership like Pam Melroy, um you know, is supportive of these kinds of collaborations. I think those kinds of
things gave me um energy. Um confidence to come into this job that people have an openness and leadership itself has an openness to these kinds of collaborations. We've got to be measured. Also, for me, it's um trying to talk about how the world, how this world we are in happened and relating internally. What I realized is sometimes I tell people at NASA stories of how we got here. You know, because sometimes even that history is not written of how this innovation happened with industry or this public private partnership.
So, I don't know that probably a long way to answer to your question. But yeah. All right here. Probably one more. Yeah. One. Oh, did you want to answer? All right. OK. All right. Well, that's it.
Well, thank you very much.
2024-02-21 05:29
