Latest / Paths of Curiosity in an AI World / Paths of Curiosity in an AI World: AI in Healthcare - People, Data, & Decisions with Mayank Goel, PhD

Transcript

Paths of Curiosity in an AI World: AI in Healthcare - People, Data, & Decisions with Mayank Goel, PhD · Paths of Curiosity in an AI World

• 0:04
  SPEAKER_00: Welcome to Pass of Curiosity in an AI world where we explore tomorrow's careers through the lens of AI.
• 0:13
  SPEAKER_00: What if your headphones could tell you're about to get sick three days before you feel anything?
• 0:19
  SPEAKER_00: Dr.
• 0:20
  SPEAKER_00: Mayant Goal is building wearable sensors that turn everyday devices into health monitors, detecting disease before symptoms even appear.
• 0:34
  SPEAKER_00: Today we're continuing our AI and healthcare miniseries with a different kind of healthcare innovator.
• 0:40
  SPEAKER_00: Someone who's never treated a patient, but whose work could change how millions of people monitor their health every single day.
• 0:49
  SPEAKER_00: Dr.
• 0:49
  SPEAKER_00: Maya Cowell is an assistant professor in the Human Computer Interaction Institute and the Software and Societal Systems Department at Carnegie Mellon University.
• 1:00
  SPEAKER_00: His research focuses on creating smarter watches, wearable devices that can continuously monitor your health without you even noticing.
• 1:09
  SPEAKER_00: His lab has developed earbuds that measure blood pressure, smartwatches that detect medication levels through your skin, and sensors that can catch early signs of illness days before you feel sick.
• 1:22
  SPEAKER_00: These aren't science fiction prototypes.
• 1:24
  SPEAKER_00: The technologies being tested today that could be in your Apple Watch or Fitbit tomorrow.
• 1:31
  SPEAKER_00: Before joining CMU, Dr.
• 1:33
  SPEAKER_00: Goel earned his PhD from the University of Washington, where he pioneered new ways to make everyday devices understand our bodies.
• 1:42
  SPEAKER_00: His work has been recognized with multiple best paper awards and has appeared in talk venues like CHI, UBCon, and MobyComb.
• 1:51
  SPEAKER_00: Dr.
• 1:51
  SPEAKER_00: Goel, thank you so much for joining us.
• 1:56
  SPEAKER_01: Yeah, thanks a lot, Acadia, for having me.
• 1:58
  SPEAKER_01: I'm looking forward to our discussion.
• 2:01
  SPEAKER_00: If your student self could shadow you for a week, what would they see you doing?
• 2:05
  SPEAKER_00: And what might surprise them the most about building wearable health technology?
• 2:10
  SPEAKER_02: Yeah, that's a good question.
• 2:11
  SPEAKER_02: So I think they'll see me sort of brainstorm a lot of new ideas with my students and talking to my collaborators who I often say I talk more to doctors at this point than computer scientists, which is interesting but also weird.
• 2:29
  SPEAKER_02: Um so I think that is something that they will see and also is a bit surprising part.
• 2:34
  SPEAKER_02: But frankly, you know, when you ask me what will they be surprised by most, the first thought that came to my mind, this is almost funny, is that they'll be surprised at how little coding I do now.
• 2:44
  SPEAKER_02: Because oftentimes what happens as a faculty member is you think of new ideas, you brainstorm with your students, but then frankly, a lot of implementation left is taken up by the students.
• 2:56
  SPEAKER_02: So I am sort of like really always at awe at what my students are able to throw together day after day, week after week.
• 3:08
  SPEAKER_00: So most of your work really is that collaboration between you, your students, and doctors.
• 3:15
  SPEAKER_01: Yes.
• 3:17
  SPEAKER_00: You don't have to be a physician to save lives.
• 3:20
  SPEAKER_00: So what drove you to create technology that does exactly that?
• 3:24
  SPEAKER_00: And how did that lead you to where you are?
• 3:29
  SPEAKER_02: So I won't say that I started with this vision that I have to convert phones into medical devices or other consumer devices into medical devices.
• 3:38
  SPEAKER_02: I was always interested in extending the capabilities of technology that we have around us to solve problems in interesting ways.
• 3:50
  SPEAKER_02: From as far as I can recall, even in middle school, the projects I used to do were always about doing more from less.
• 3:59
  SPEAKER_02: Instead of learning something new, what I've always wanted to do was whatever I know so far, what's the best way I can use this to solve an interesting problem.
• 4:07
  SPEAKER_02: So it was with that vision I went to grad school.
• 4:10
  SPEAKER_02: And then my advisors there at University of Washington were already working on some aspect of health and some aspect of helping people in the developing world or other resource-constrained environments.
• 4:24
  SPEAKER_02: Combination of those three passions of working on health, helping people in need, and doing more from less meant that I was on this path that I've charted for the last 15 years.
• 4:37
  SPEAKER_00: I love that.
• 4:38
  SPEAKER_00: Doing more with less.
• 4:40
  SPEAKER_00: Your lab builds earbuds that measure blood pressure from just behind the ear.
• 4:45
  SPEAKER_00: Can you describe that journey from something that sounds impossible to actually making it work?
• 4:54
  SPEAKER_02: Yeah, it's a good question.
• 4:56
  SPEAKER_02: So frankly, I won't say it just works even now, right?
• 5:00
  SPEAKER_02: Because all of these are research prototypes that we test out.
• 5:04
  SPEAKER_02: But what happens is even if it works really, really well, for an end user to use it beyond our studies is a really, really high mark, and justifiably so.
• 5:14
  SPEAKER_02: So getting there, there are still some steps we have to take.
• 5:17
  SPEAKER_02: I think this also started when I was a grad student at the University of Washington, and my advisor came to us and said, Hey, do you think this is something we can do?
• 5:25
  SPEAKER_02: And not on earbuds or anything, but just using devices that we have around us.
• 5:29
  SPEAKER_02: And all of us started brainstorming, we thought of different ideas.
• 5:33
  SPEAKER_02: And it's interesting that the three or four students who were thinking of how we can solve this problem are either now researchers in big industrial labs or faculty members at big universities in North America, and everyone has their take on that one problem.
• 5:53
  SPEAKER_02: And that is just one example, right?
• 5:55
  SPEAKER_02: That you have a problem that someone brings to you.
• 5:57
  SPEAKER_02: Like my advisor was also not the first person to bring it to us at that point.
• 6:00
  SPEAKER_02: He talked to a doctor who introduced him to this problem.
• 6:03
  SPEAKER_02: And I remember that doctor because we worked together on another health sensing project, and he said that if you can give me 24-7 measure of blood pressure, I am certain we can cure most of the illnesses that humans face.
• 6:18
  SPEAKER_02: So because we knew that okay, there is a champion in the field, a practitioner who believes that there is value in solving this problem.
• 6:26
  SPEAKER_02: We didn't have a solution at that point, but we had some interesting ideas, and then we just ran with that all in our own different ways.
• 6:33
  SPEAKER_02: And yeah, it has been a journey.
• 6:36
  SPEAKER_00: So it's a project you've been expanding on for quite a while.
• 6:42
  SPEAKER_02: Yes.
• 6:43
  SPEAKER_02: What happens is you know, sometimes you start with a small idea and it doesn't work in that moment, but you keep that problem with you because you are excited about the problem, but you still don't have a solution.
• 6:53
  SPEAKER_02: So it doesn't mean that I've been working on these things for the last 15 years.
• 6:56
  SPEAKER_02: A lot of times you just keep the problem in your back pocket.
• 6:59
  SPEAKER_02: And when some interesting technical breakthrough or a Eureka moment happens, either to you or your collaborators, you like bring it back.
• 7:07
  SPEAKER_00: Right.
• 7:08
  SPEAKER_00: So some students might think my smartwatch already tracks heart rate and steps.
• 7:13
  SPEAKER_00: Isn't that enough?
• 7:15
  SPEAKER_00: What could next generation wearables do that current devices cannot?
• 7:19
  SPEAKER_02: So I can think of two high-level things that are missing right now from smartwatches.
• 7:24
  SPEAKER_02: One, whatever the smartwatches measure right now, we frankly do not have a good use of that.
• 7:32
  SPEAKER_02: No one really looks at the heart rate information on their watches apart from an athlete who's trying to figure out how many calories did they burn during a workout or things like that.
• 7:44
  SPEAKER_02: And the step count as well, I feel that for the most part, people use it initially to figure out how active they are.
• 7:52
  SPEAKER_02: They try to use it for a bit of a behavior change.
• 7:54
  SPEAKER_02: But once they build a mental model of their behavior patterns and how much they are walking, people stop looking at that number.
• 8:03
  SPEAKER_02: Most people stop looking at that number.
• 8:05
  SPEAKER_02: So I feel like the things that a watch can already measure, we are not using them to their potential even now.
• 8:12
  SPEAKER_02: So that is one.
• 8:13
  SPEAKER_02: And second, I think there is no end to what all we can measure on these devices.
• 8:19
  SPEAKER_02: Whether it is your smartwatch, whether it is your smartphone, they are by default designed to mimic the human senses.
• 8:28
  SPEAKER_02: And humans are able to do so much with their senses.
• 8:31
  SPEAKER_02: So these sensors can also do a lot, but what I need is different from what you need and what the third person needs.
• 8:38
  SPEAKER_02: I think with the advent of modern AI algorithms, and as AI starts sort of writing better and better code, I think we'll start seeing applications that are designed specifically for me.
• 8:51
  SPEAKER_02: The app I have on my watch helps me take care of my health in the way I and my doctor wants.
• 8:58
  SPEAKER_00: So we have all this data and we're barely scratching the surface.
• 9:02
  SPEAKER_00: You've mentioned AI a few times.
• 9:05
  SPEAKER_00: How do you use it to truly add value?
• 9:08
  SPEAKER_00: And where is it more trend than necessity?
• 9:12
  SPEAKER_02: So I'll answer the second part first that how is it a necessity?
• 9:16
  SPEAKER_02: So the example that I gave you of blood pressure, where a doctor, collaborator came to us and said this is a problem worth solving.
• 9:24
  SPEAKER_02: Every problem related to health sensing is a problem that came to us through a practitioner.
• 9:31
  SPEAKER_02: So we always make sure that we are working on things that the practitioners, the experts in the field believe is worth solving, and then we solve it.
• 9:41
  SPEAKER_02: Actually, in some cases, we come up with the idea as well initially, but then before we even build that initial thing, we go to the doctors and we tell them that this is what we are thinking.
• 9:50
  SPEAKER_02: Do you think it makes sense for us to do?
• 9:53
  SPEAKER_02: So, for example, we are right now working with children who have hyperactivity.
• 9:58
  SPEAKER_02: And one of the solutions that we wanted to build was using AirPods and smartwatches and smartphones to identify how the human body is moving.
• 10:07
  SPEAKER_02: These devices can already count our steps.
• 10:09
  SPEAKER_02: But my students have built algorithms that can go much beyond that and also see how the body is shaped at any given moment.
• 10:17
  SPEAKER_02: How am I sitting right now?
• 10:18
  SPEAKER_02: Am I sitting cross-legged?
• 10:20
  SPEAKER_02: Am I hunching?
• 10:21
  SPEAKER_02: What are my hands doing?
• 10:22
  SPEAKER_02: All of those things with just these devices.
• 10:24
  SPEAKER_02: So we had that algorithm.
• 10:26
  SPEAKER_02: We had not used it on children.
• 10:28
  SPEAKER_02: But the idea we had was that children who have hyperactivity, would it make sense to try some of these technologies to have our algorithms that we have built for grown-ups on small children and provide a doctor a window into the child's life that you know their math teacher keeps saying they are inattentive, they are hyperactive.
• 10:50
  SPEAKER_02: Can I provide the doctor a window into the child's behavior in that math class and see what really is happening?
• 10:58
  SPEAKER_02: But without putting cameras, because I do not want to invade privacy of everyone involved.
• 11:04
  SPEAKER_02: So we had that idea.
• 11:06
  SPEAKER_02: We took it to our collaborators who we have been working together for almost a decade at this point.
• 11:11
  SPEAKER_02: They were like, this is a cool idea.
• 11:13
  SPEAKER_02: And we started collecting data.
• 11:14
  SPEAKER_02: We still don't have results, but I'm just telling you that as an example of how these things sort of take shape.
• 11:22
  SPEAKER_00: Too true.
• 11:23
  SPEAKER_00: So as someone who is not a physician, how do you know what kind of problems you should actually be focusing on?
• 11:33
  SPEAKER_02: So, Sandy, there is no one answer.
• 11:35
  SPEAKER_02: A lot of it is organic, and a lot of it is based on brainstorming that you do with your colleagues.
• 11:40
  SPEAKER_02: So when I talk to my doctor collaborators, they introduce us to new ideas, they tell us things, and again, we keep those ideas at the back of our mind.
• 11:48
  SPEAKER_02: And you read research papers, you read articles, and one of the main things we do, frankly, is connecting the dots.
• 11:57
  SPEAKER_00: When a project doesn't work, what is usually the biggest challenge?
• 12:01
  SPEAKER_00: The sensors, the algorithms, the users, something else?
• 12:08
  SPEAKER_02: Usually the biggest challenge is the sensors or the data that we need to collect.
• 12:14
  SPEAKER_02: That sometimes the problem is hard enough that you need data from lots and lots of people.
• 12:19
  SPEAKER_02: And it is not easy to do.
• 12:21
  SPEAKER_02: For example, a lot of advances in machine learning are built on data sets that the communities have collected together.
• 12:28
  SPEAKER_02: That is something which is really hard to do when you're working on health sensing on medical-medicine-related projects because the data is inherently private.
• 12:37
  SPEAKER_02: So you will see that there are research labs across the world working on very similar problems, but very little data gets shared between them.
• 12:46
  SPEAKER_02: So that frankly ends up being a huge problem.
• 12:50
  SPEAKER_00: That privacy piece is something I wouldn't have expected to be such a bottleneck.
• 12:55
  SPEAKER_00: For students who want to build health wearables, what should they actually be learning right now?
• 13:01
  SPEAKER_00: What skills do you wish more students came in with?
• 13:05
  SPEAKER_02: I think this is an answer that the Nvidia CEO gave a couple of months ago, but frankly, that is the answer I'll give as well.
• 13:12
  SPEAKER_02: Physics.
• 13:13
  SPEAKER_02: I would love if the students spent more time learning physics and you know, maybe even chemistry and biology, but I was not great at chemistry and biology, so I can relate less to that.
• 13:23
  SPEAKER_02: But definitely basic sciences is the most important thing because those are the things that don't change, those are the principles of our being here.
• 13:31
  SPEAKER_02: So what happens is a lot of sensing that we do is about making that connection between two physical modalities.
• 13:39
  SPEAKER_02: So if you know that these two things are related because of what you learned in high school, gives you that confidence that with enough data, with enough processing of that data, with machine learning, I will be able to find a relationship between these two things.
• 13:55
  SPEAKER_00: So it's really about understanding the fundamentals, not just chasing the latest tools.
• 14:01
  SPEAKER_00: For a student listening who wants to build health technology as an engineer, what can they do now?
• 14:06
  SPEAKER_00: And what should they focus on over the next year or two?
• 14:12
  SPEAKER_02: I think it is very important to keep up to date with the modern advances in technology.
• 14:18
  SPEAKER_02: I have hinted at the use of AI a few times already on the podcast, and I think the world is moving at a breakneck speed.
• 14:27
  SPEAKER_02: So being exactly at the forefront of it is just really tiring.
• 14:31
  SPEAKER_02: I don't think any human can do this at that point.
• 14:33
  SPEAKER_02: But in our own ways, we need to remain abreast to some of the advances so that you know what others are working on, and ultimately the world is built by standing on the shoulders of giants.
• 14:43
  SPEAKER_02: So I think we just need to keep doing that.
• 14:46
  SPEAKER_00: Keeping up without burning out is its own skill, honestly.
• 14:50
  SPEAKER_00: Technology is evolving so fast.
• 14:53
  SPEAKER_00: Over the next 10 to 20 years, what do you think will change most about wearable health technology?
• 15:00
  SPEAKER_00: And what new opportunities might it create?
• 15:05
  SPEAKER_02: As I mentioned earlier, I think like we will start building solutions that are tailored to individuals.
• 15:10
  SPEAKER_02: Right now, we measure things that the big corporations are able to build systems to do and what the doctors believe we need right now.
• 15:17
  SPEAKER_02: But the challenge is that the doctors are also trained on the data that we had 15 years ago.
• 15:22
  SPEAKER_02: Building a mental model of what we can measure right now is not easy.
• 15:26
  SPEAKER_02: Right?
• 15:27
  SPEAKER_02: We can already measure blood pressure 24-7, we can measure breathing 24-7, however long someone sleeps, like the whole time, heart rate, all these things.
• 15:36
  SPEAKER_02: How is that data useful?
• 15:37
  SPEAKER_02: What does that mean for the standard of care?
• 15:40
  SPEAKER_02: Are still unanswered questions.
• 15:41
  SPEAKER_02: So I think as we answer these, as we figure out that what is the right use of these systems, we'll realize that all of us are needs are slightly different.
• 15:50
  SPEAKER_02: And you know, companies like Apple, Google, Samsung, Microsoft, these companies are not necessarily going to build solutions for one person.
• 15:58
  SPEAKER_02: They'll build solutions that will help billions of people.
• 16:02
  SPEAKER_02: But for that one person, I think that's where the modern advances will end up proving much more useful.
• 16:09
  SPEAKER_00: Solutions built for one person.
• 16:12
  SPEAKER_00: That's a future most people haven't even imagined yet.
• 16:16
  SPEAKER_00: If you could go back and talk to yourself as a student, what is one piece of advice you would give?
• 16:25
  SPEAKER_02: That's a good question.
• 16:26
  SPEAKER_02: I actually do not have a good answer apart from saying build bridges.
• 16:31
  SPEAKER_02: One of the things I said earlier is that I always challenged myself to do more from less.
• 16:37
  SPEAKER_02: I feel I might have overdone that a bit.
• 16:41
  SPEAKER_02: One message I would give to myself is build more bridges, learn more from others, and then see when you do that, you are able to amplify the kind of impact you can have, the kind of solutions you can build.
• 16:51
  SPEAKER_02: And that is something that I learned more in grad school.
• 16:54
  SPEAKER_02: And after coming to CMU, I would have ideally liked to learn that slightly sooner.
• 16:59
  SPEAKER_00: Build more bridges.
• 17:00
  SPEAKER_00: I think every student needs to hear that.
• 17:03
  SPEAKER_00: What is the best career advice you ever got?
• 17:08
  SPEAKER_02: Keep learning.
• 17:10
  SPEAKER_02: And if people don't believe in your idea right now, keep trying.
• 17:14
  SPEAKER_02: At some point, you'll be able to convince others of the value of your idea.
• 17:19
  SPEAKER_02: And on a similar note, one more thing that comes to my mind that my advisor told me that if you ask someone and they say no, go ask someone else.
• 17:27
  SPEAKER_00: So simple, but so hard to actually live by.
• 17:31
  SPEAKER_00: Is there any project you have in mind right now that people might think is impossible, but you strongly believe in, anyways?
• 17:41
  SPEAKER_02: I think building a custom medical device for each human, that that human is able to build by themselves, is something that people don't believe right now.
• 17:53
  SPEAKER_02: So, for example, a lot of people say that AI will take a lot of jobs that humans do.
• 17:59
  SPEAKER_02: There are reasons to believe that that might happen, but I think one thing that we are forgetting is we will also have automated ways where everyone's problems get solved at an individual level.
• 18:11
  SPEAKER_02: Right?
• 18:12
  SPEAKER_02: So I think solutions that work for n equal to one, not only in the digital world, but also in the physical world, is the future that I'm excited about.
• 18:21
  SPEAKER_02: But I think not everyone is convinced by that yet.
• 18:25
  SPEAKER_00: Finally, are there any books or podcasts you'd recommend to students who want to dive deeper into this space right now?
• 18:36
  SPEAKER_02: I think podcasts like yours, I think listening to young folks, young readers, young content creators is very, very important.
• 18:44
  SPEAKER_02: I recently was listening to a podcast or an interview by Barack Obama where he said that a lot of world problems are because older people refuse to sort of leave, or like something like that.
• 18:56
  SPEAKER_02: I'm paraphrasing him.
• 18:57
  SPEAKER_02: I think we need to do the opposite of that.
• 18:59
  SPEAKER_02: We need to learn from younger generation, younger people, because that is the best way of like keeping up with the latest knowledge and trends.
• 19:09
  SPEAKER_00: Dr.
• 19:10
  SPEAKER_00: Goel, thank you so much for showing us what it looks like to engineer the future of health monitoring.
• 19:17
  SPEAKER_00: For students listening, remember the goal isn't to find that one perfect major, it's to build skills at the intersection of your interests, whether that's engineering and health, AI and design, or hardware and human behavior.
• 19:34
  SPEAKER_00: If today's conversation sparked your curiosity, explore the other episodes in our AI and healthcare mini-series.
• 19:41
  SPEAKER_00: From emergency room positions using AI to save lives to pediatric informaticists designing AI for children.
• 19:51
  SPEAKER_00: And to help you visualize how these careers are evolving, check out the AI and Healthcare Future Jobs map and resources linked in the show notes.
• 20:00
  SPEAKER_00: This has been Paths of Curiosity in an AI world, where we explore tomorrow's careers through the lens of AI.
• 20:08
  SPEAKER_00: Until then, stay curious.