Trust, Data, and Decision-Making Beyond Earth — Andrew Feustel, Former NASA Astronaut

0:00One of my most amazing memories from space was the first instant that I saw Earth from space, right? That's a memory that you never lose because as a human, you hope that someday every human gets to see it. But as a human who's seen it, you'll never forget that the microsecond instant that you had that view of Earth from space.

0:24Welcome to Data & AI Mastery, the podcast where we bring you cutting-edge insights, practical advice, and inspiring stories from the leaders shaping the future of data and AI across the globe. I'm your host, Raoul Gabriel-Urma, founder of Cambridge Spark, the leader in transformational data and AI upskilling, career development, and progression. In each episode, I will be diving into real-world case studies of companies harnessing the power of AI to drive innovation, reduce costs, and create new business opportunities. So whether you are an aspiring data scientist, AI engineer, or seasoned executive, this show is designed to give you the tools and knowledge to stay ahead in a world where data is transforming every aspect of business.

1:09Stay ahead, stay inspired, stay masterful. Welcome to Data and AI Mastery. Hi, Drew. How are you? I'm great, Raul. Great to see you. Thanks for having me on the show.

1:27So, Drew, I'd love maybe to start just a little bit your story and how did you become an astronaut? Well, there's probably a longer answer to that question. Um, but, uh, you know, the simple answer is that I applied to NASA's selection process and I did that in the, uh, basically in late 1999. So showing my age, I was an astronaut for 23 years with NASA, uh, joined at the age of 35. And, uh, it was, uh, I I'd say it's a simple process. It's a matter of, uh, putting an application and it's quite a long application and, uh, the road to getting from application to actually being selected is, is, uh, not trivial, but, uh, I'm a, I'm what's considered a scientist astronaut.

2:17I didn't go through the military. I'm not a test pilot. Um, I am a pilot, but not, uh, a jet aircraft test pilot. And so, uh, you know, you fill out an application and then you cross your fingers and hope that you make it to the final selection process. Amazing. And you've been to space three times, if I'm not mistaken? Yeah, three times. Twice on the space shuttle and once on a Russian Soyuz rocket to the International Space Station. I had great privilege on my first flight in 2009 aboard space shuttle Atlantis to launch to the Hubble Space Telescope and perform during the final servicing mission by astronauts of the Hubble Space Telescope, although we could always go back again with the right set of resources.

3:10And then in 2011, I flew on Space Shuttle Endeavour, the final flight of Space Shuttle Endeavour, to the International Space Station for 16 days to do some work. We called a utilization and assembly flight, which means we completed the assembly of the space station and took up some hardware and supplies and took a few things home on that mission. That was 2011. And then seven years later in 2018, I launched to the International Space Station for 197 days and worked on, it was my chance to actually get to live in space as opposed to my previous two flights where I found I was just working or visiting space.

3:53This time, I got to go to stay and led the team on the ISS. I was commander for, I guess, about five and a half of the six and a half months on the space station. And that was my most recent flight with NASA.

4:16I guess to start with, you know, in the business world, if you get your data wrong, it can cost you a lot of money and you're going to, you know, make decisions that affect the business performance. I'd love to hear, you know, when you're in space, like what data do you have access to, you know, as an individual and, you know, like what do you do with it and, And, you know, why is it so important in that world? I mean, data is the key to everything that we do. And I think there's, you know, there are differences between the business world and, in my experience, the human space exploration world.

4:55And in human spaceflight, we have access to many different types of data. Those related to flying the spacecraft, those related to, you know, data related to experimentation on board the space station or our orbital platforms, and data related to commanding and control of the space station itself as it's orbiting the planet. So, you know, data is everywhere for us. NASA and the space community relies on data from satellites and telescopes, space telescopes looking out into space for, you know, many things even unrelated to human space exploration. But in all of the analysis that's done, you know, data is kind of the, I mean, it's the root of everything we do.

5:48From a real-time setting, I'm just imagining you in space doing the repair and you look at telemetry data and you're kind of thinking, this is not matching up maybe your reality of what you're seeing or what you're doing right now. Was there any situations like that that you've come across or it's always data meets reality perfectly? No, data never really matches reality. But, you know, it's interesting to think about the concepts of data for an astronaut and human space exploration. Like, clearly data plays a role in everything that we do working in space, whether that's actual humans working or with the data we collect.

6:29And there's sort of two different worlds. And I thought, you know, before the program, you know, you sent out some information about some of the things we might talk about. So I started thinking about those things and how does how does data, AI, you know, play into the work that we do as astronauts. And the reality is that to a large degree, it doesn't yet. And I think the reason is because a lot of the work that we do is is hands on, like physical work. A lot of the repair, you know, my my spaceflight career has been focused largely on spacewalking and external repairs to spacecraft. So I was never was never the pilot of the space shuttle.

7:17I was never the commander. My my role is always in spacewalking and going out and doing external work. So from my perspective, that data was really the data that I see in my eyes, you know, my perception of the things around me and my reaction to the environment around me. As we've talked about exploration of planetary surfaces and some of the techniques and methods we'll use to do those things, we have incorporated and identified the need for data to be provided to us. So, for example, something that we don't have on our spacesuits right now for external work on the space station is we don't have like a heads-up display in our visor providing, you know, temperature, pressure, spacesuit information, any information about the surroundings or distances to objects.

8:08We don't have any graphical data that could be fed to us by researchers on the ground about areas of interest that they might see in the foreground. But all that information could be really useful to an astronaut on scene on a planetary surface trying to go investigate something. And so that's an area where we'll start to roll data into our physical system, the world that we live in, to be able to process that in the tasking that we do. There's certainly data that's associated with flying a spacecraft, and we have a requirement as pilots of those spacecraft to synthesize that information and then react to it or react with it to make sure we're meeting our target objectives, whatever that is, with the spacecraft at that time.

8:58So all critical, but so many facets to the ways that we incorporate data into the work that we do. It's very situational dependent. that's super fascinating so it feels like there is a clearly data involved for the planning of the mission uh making it safe uh but then there's also so there's that to talk about but then there's a space where you know you as an astronaut you're doing the work then uh actually if i'm hearing right fundamentally that world is not data rich yet as a user uh it's perhaps collected for further analysis later on, but as a user, you're not actually, you're kind of having to rely on your human inputs.

9:42Yeah, you kind of do. And as pilots and astronauts, we always struggle with the engineers and the designers to ensure that there is manual control systems on our spacecraft or ways for us to manually override the systems because we don't yet have the trust in the systems because of the criticality. So if you're relying on an automated system to toast your bread and you use a voice command that says, make me toast, and you know that the whole system and the process will be executed based on that voice command, and out comes the toast, and it's cooked to the degree that you expected it to be, well, then that builds trust, and as you do that over and over again, you know that when you say that thing, then something results.

10:30But it's not a critical system. But if you say, OK, system, land me on the surface of the moon in a boulder field, you may not achieve those objectives and we may not trust that the system can do that. So our ability to utilize those is somewhat limited now, and that's going to change as we get exposure. So we create that learned trust in the system, which will allow us to sort of lower our guard a little bit on, you know, that situational criticality. And I think that's sort of key to why we don't see as much data-driven and automation processes in our human critical systems yet because we haven't evolved them all.

11:22I hope you're enjoying today's conversation. If you're finding the insights useful, please do take a moment to subscribe to the Data and AI Mastery podcast and leave us a review on Apple Podcasts, Spotify, or YouTube. Every new follow helps us reach more people and shed incredible work being done by today's data and AI leader. All right, let's go back to the episode. That's fascinating because this idea of trusting the system autonomously, overriding, is a big topic with artificial intelligence right now. Do you trust AI to produce the right result or take the right decision for you or give you the right diagnosis?

12:01So this idea of trust is real in the business world. It sounds like what you're describing here really depends on the criticality. And because in space, wrong decisions are catastrophic, that's driving the bias towards human decision instead. That's really interesting to hear. So I guess, what would it take? And do you think it's even possible for all decisions to be completely automated away while in space? Is there a world where we can get there or fundamentally space exploration is a human task? And I think automation and AI systems in telescope or sensor, you know, spaceward looking sensors and data collection, that's certainly relying on AI.

12:54because as we put more advanced, as we turn more advanced systems out into space to collect data, there's just such massive amounts of information coming back that it's, we really don't have the human capital to parse through all that data. And there's really no need because much of it is probably data that we don't need to pay attention to. I mean, it's all important, but may be not important for the reasons we're looking. So if we're looking for planets like Earth, there's a certain set of criteria that we would identify that a particular sensor would respond to. And so we feed that data into the system, it starts acquiring information.

13:35And then the system itself that's acquiring data has to decide whether it meets the flags that we've set that say, hey, this is a planet versus not meeting those flags. And so it's really important to have those AI and automated systems there to process those massive amounts of data that are coming back from these telescopes. For that, we probably couldn't achieve what we have in the most recent decade or two in terms of discoveries in space without AI helping us to, or data processing systems helping us to understand, you know, what it is we're looking at and sorting that data into things to pay attention to and things that we don't need to right now.

14:25Great. So it's really basically, if I play this back, the main use case here is we can't scale human capital. So it's kind of like there's no choice but actually leverage AI and robotics, at least in space, to crawl through the data itself while it's there because we can't send that many people in space to do it anyway, and that would probably be a bad way to do it. Well, we certainly can't send them very far either. I mean, right now the moon is about as far as we can get. We can get to Mars, but it takes a little bit longer, and we don't really have human support systems. Keeping humans alive in space is like, that's the big thing.

15:04You could argue that, well, then we should just send robots to space to do all the work that humans would do, But I would argue that we've sent robots to space and we've got multiple rovers on the surface of Mars over the last, you know, two decades, two and a half decades. And although they've been incredible at investigating the surface of the planet and all the conditions that are there, the amount of data and the distances that those rovers typically cover over the course of, you know, five years can be achieved by a human within just a few months. So part of exploration is about speed and ability to collect information quickly.

15:47And rovers are pretty good, but they haven't reached the point yet where we can just set them free and have them go off and do things quickly. It just takes time. And they don't make fast decisions because we don't want them to break themselves in the work that they do. That's really interesting. So there's clearly still things that humans can do better and faster and there's more sense to be involved versus the, I guess, the rover and robotics side. That's really interesting. So I guess maybe I'd love to take a step back and think about what does it take to get to space, you know, broadly, you know, NASA's organization is obviously like hugely inspiring.

16:28So like walking through, planning the mission, preparing the people, having the, you know, the right data to decide when, you know, we're going to go to space and so on. Out of everything you've seen through your career, you know, in terms of like processes, in terms of like culture and mindset and thinking about what's important, what do you think is what makes NASA so special on that side and maybe that businesses could learn from? Yeah, so I would say, you know, NASA really historically has been the gold standard for preparing humans to go to space. And that's because NASA has always led the way in doing that in human spaceflight preparation and other sovereign nations and now individuals follow the standards that NASA has set in preparing humans to go to space.

17:21And as you said, it does take a long, long, you know, a lot of preparation. It takes years of preparation and planning to be, you know, to get selected in the first place. And then missions themselves, human preparation is anywhere from, you know, at minimum eight months and even up to three years or so just to prepare for a single mission. for the Hubble Space Telescope mission, which was 13 days total in space, we prepared, we trained for three years for a 13-day mission. That's a long time, and that's a huge investment in human capital, not only the people that you're sending to space, but the team that's involved in training the individuals to go.

18:06So I think the lesson there for businesses is that great success and high risk and high performance takes a considerable investment in human capital and skill building to achieve those goals. And I think it may be short-sighted to think that you can achieve such positive objectives and such broad and significant objectives without investing in the training and the expertise that you need to really have people be experts in the things that they're trying to move forward. I guess key innovations that maybe are missing today, or how do you think AI can help us achieve that vision you described quicker?

18:54I think it already is. So a good example is SpaceX, the company, right? They've managed to... to SpaceX had a contract with NASA to provide cargo to the International Space Station and eventually now provides transportation to the International Space Station with the SpaceX Dragon. And what's unique about that spacecraft is that it has a reusable first stage segment. And you've seen it, I'm sure you've seen it landing back on the surface of the Earth. Space shuttle was reusable as well, but not many of the rocket components, the fuel tank that we used and the solid rocket boosters. Those were sort of reusable, but not nearly the same way.

19:41But that spacecraft can come back and land with automated systems and commanding, not from humans controlling it or flying it back, but because of AI and automation and the ability for that rocket to guide itself back to a landing platform and control all of the systems that it needs to to land that back on the surface of Earth so that we can reuse it. And so AI is already playing a huge role. It's interesting to use the word AI. I mean, AI is just a term we've applied to what we've been doing all along, which is essentially fusing sensor data, you know, taking information in from all of our sensors that we have on these systems and all of the components that are controlling the propulsion and some of the aero surfaces on the spacecraft to guide it back and the landing legs, all those things, right?

20:36Fusing all that information together and then making next step decisions based on the information that's coming in with the target objective of vertical landing at a certain place. So anyways, that's been very effective, and that will be applied towards the next phase, which is landing on the surface of the moon successfully with humans. And all of those things related to navigation and guidance as you're approaching the surface of the planet. If we think back to the Apollo era, there have been six splites to the surface of the moon. Every single one of those splites, though they started off with sort of automatic landing systems, for every single flight, astronauts took over for the final landing.

21:19And primarily because the systems were not designed well enough with enough information to understand that what the astronauts could see on the surface was a much higher risk level than what the system was perceiving. And so as we move forward in our attempts to land repeatedly and successfully on the surface of the moon, we will need to automate those systems and create a decision-making process, which is basically AI that says, I'm including all of these sensors and data into my control systems, and I'm placing the rocket, you know, in a location on the surface that will keep the astronauts alive when we land and not destroy the vehicle.

22:04And so, you know, that's sort of the next phase is incorporating all the systems we're building to put rockets back on the surface of the moon or on Earth into those rockets that are going to land on the surface of the moon with much greater hazards because they're not prepared landing surfaces like what we have on Earth. It's super cool because there's so many... Maybe we would get into all these details. I don't know. Hey, like, I'm amazed by what you said, the six Apollo missions in all scenarios the crew took over I had no idea so that's fascinating that you know and the key lesson is well the data that the model was trained on wasn't representative of the reality and that's quite common in the business world you implement a model and use your training data and then in reality you know there might be a different pattern completely and COVID was an example right like pre and before traffic in streets was very different so I guess it's kind of interesting nugget

23:05hey Drew I'd love to take you to maybe a couple of quickfire round of questions okay there's one I was asking my friends you know like what what question can I ask and this one's really good is what's the memory that you'll keep for yourself forever from the international space station like just being there what's like one memory you have yeah it's uh the thing is there's not one. You know, when you ask that question, like just the fact of you asking that question immediately in my mind starts just coming like multiple events, you know, from every mission, like those highlight things that I remember, I can tell you a few of them.

23:41Like one of my, one of my most amazing memories from space was the first instant that I saw earth from space, right? That's a memory that you will, that you never lose because as a human, you hope that someday every human gets to see it. But as a human who's seen it, you'll never forget that the microsecond instant that you had that view of Earth from space. And for me, I think we were over the Himalayas when I got to look out on Earth and, you know, just to experience what that looked like from, you know, 300 miles or 500 kilometers away was just incredible. So that's one for me. I have some moments from spacewalking that I remember, you know, flying over Texas outside on a spacewalk in my suit, holding on to the side of the Hubble Space Telescope and looking down to see Houston, which was my current home with my family and everybody there.

24:40and being able to look and cross the entire United States all the way to Michigan, which is easy to spot from space because it looks like a hand. So I could see my home in Texas and I could see entirely across the planet to see Michigan and the place where I grew up and Ontario, where I went to school, where my wife's family's from. That was a really special moment for me. And, you know, all the while I'm just hanging with one hand off the side of this telescope looking between my legs. I mean, that was really a unique perspective to have. So there's something on every flight that you never forget.

25:22Well, thank you for sharing. Just living it through by hearing you. So that's amazing. Two more quick questions. That's cool. What was your favorite subject? Maybe surprisingly or not surprisingly, I wasn't very good at them, but I always enjoyed physics and I always enjoyed chemistry. I wish I was better at both of those things, which sounds funny coming from a guy who has a Ph.D. in physics. But I had challenges in school. I attribute my wife to my successes in university because she gave me an ultimatum that was my wife, Indra, said, And you can either hang out in the library with me or you can go date somebody else.

26:05But that's where I'll be. And that kind of got me to sort myself out and do a little bit better in school. So, you know, here I am. But anyways, physics and chemistry were two of the things that I really enjoyed when I look back on it. Amazing. And final question. What's your favorite music genre? Rock and roll. I mean, I'm old enough that I grew up with rock and roll and still to this day. Yeah, you know, Led Zeppelin, all the greats, The Who. One of my favorite bands is The Tragically Hip. It's a good Canadian rock band with some strong sort of roots in good old rock and roll. So I have a musician as well.

26:46I'm not a very good one, but NASA's astronaut office had a band called Max Q. It's been around since 1986. And so I played in the band for about 20 years as a guitarist and played some music in space, even recorded a song up on the space station, which is kind of cool. And the song was written by one of my friends, Gord Sinclair and the Tragically Hip. And so, yeah, I like music. I like rock and roll. And, you know, music is the spice of life. Thank you, Drew. It's been a great pleasure to have you on the Data and Air Mastery Show. You're welcome, Raul. Thanks for having me. And I hope we got some good, created some things for people to think about today.

27:35I really enjoyed this episode with Drew. I mean, what an inspiring person and what a privilege to have this conversation. Clearly, the world of space is rich of data. Now, there's a couple of really interesting nuggets. Now, he explained how autonomous decision systems and AI can help rockets land back on Earth often as a result, you know, that drives the cost down to ship more into space. So that was a good example because it shows how AI can help innovation and can support more space exploration. Now, on the other side, what I thought was maybe surprising and contrasting is that there's also a bunch of scenarios where you do not want to use autonomous system because we don't yet trust the algorithm or the machine, right?

28:25So one example was when the Apollo mission was landing on the moon, in all the scenarios, the astronaut decided to override the system because they could see that actually we're dealing with new data here and it doesn't know what to do. So we can't just trust the autonomous system completely, especially when it's associated with a high criticality. You do want to have that override system. you do want the human to be able to take control. And I think we can really resonate with that in the business world, right? We have AI systems that can take decisions about your customer, can, you know, with agents now, send an email and so on.

29:11In some scenarios, you want to trust AI to do it. but if it's high criticality especially in environments that have high regulations like healthcare for example or finance then actually you might well want to have a manual process and have a strong community loop so I think that was very interesting Thank you for tuning into this episode of Data and AI Mastery If you found value in today's discussion make sure to subscribe so you never miss an insight from the leaders driving the future of data and AI. And if you're a data and AI leader looking to upskill your workforce with the fundamental data and AI skills to transform your business, Cambridge Spark is here to guide you every step of the way.