George Netscher on The Robot Brains Season 2 Episode 10

 

Transcript edited for clarity

 

Pieter Abbeel: We often talk about the hope that one day artificial intelligence will make our lives and the lives of our loved ones safer. But all too often, it can feel like something for the future. But today's guest is already building an AI to help vulnerable people right now. Motivated by his own life experiences, George Netscher is pioneering AI technology to help the elderly and others with dementia be safer in their homes. SafelyYou uses artificial intelligence-enabled vision systems in the fight against dementia. While still a young company, SafelyYou, has successfully reduced falls by 40 percent and E.R. visits resulting from falls by 80 percent. Welcome to the show, George. We're so excited to have you here. 

 

George Netscher: Thanks for having me, Pieter. Super excited to be here. 

 

Pieter: George, the first time we met was, of course, in the Berkeley AI Research Lab while you're working on your Ph.D.. Why did you get into doing it and how did that lead you to starting this company? 

 

George: Yeah, that must have been about seven years ago now. Pieter was just a young faculty at that time. Yeah so you could go back and look at my graduate application essays from back then, and this was in 2014. It was all about how can we use these new tools in A.I.? You know, everybody was saying we passed this inflection point with Alexa in 2012, and there has to be some way that we can use these tools to help my family. And it was really just about how do we help people with Alzheimer's disease? And I didn't realize how much of a weirdo I really was actually coming into a computer science based program, but I had such an application in mind that I really wanted to work specifically on, you know, leveraging the tools to help people with Alzheimer's. But the opportunity just seemed so clear and that there has to be some way that we can give a voice to a population that loses the ability to advocate for themselves through tools like this. So that was really the dream from the beginning, and I like to say that I'm in year 20 of the 30 year plan, but it really still feels like just the beginning that there's so much good that we can do going forward. 

 

Pieter: And that's so interesting. A lot of people coming to their Ph.D. thinking purely about technology. But here you are saying from day one, you already had a very specific, important application in mind that you thought doing the Ph.D. would help you solve that problem. Is that right? 

 

George: Yeah. And. Yeah, 100 percent, and that it felt like the way to push things forward was by, you know, deeply understanding what the latest and greatest is and doing something that didn't seem technically feasible. And, you know,pushing the boundaries of how we could help instead of, you know, just repeating what was already out there, whatever else that there had to be a way to leverage these new tools to really do a lot of good for folks. Then obviously, given so much of the power of these tools comes from the data you have. And it became a bit of a journey of, “Oh, man, you have to be really crazy to approach a problem like this with data that is so hard to get and it's so, you know, so important, but so private and personal and and all of that that”, you know, sitting next to the folks that we're working on cafe in the early days and things like that that I was the crazy one that wanted to go spend all my time with old people and see if I could, you know, find ways to support them now. 

 

Pieter: Dementia. Can you say a bit more about how many people are affected by it and how does it manifest itself? 

 

George: Yeah, definitely. And this is stuff of wanting to help my own family and I saw how challenging it was. So in my family, for reference, my mom's mom had it and is now passed. And then my mom's big sister has it, and it's now late stage, and then my mom is basically the next one of mine. So I saw how challenging it was for her, but also felt very kind of inevitable that mom is the next one. So really, the goal is how do we build what I want for my own mom before she gets to that point? When we were first starting, it felt like it was maybe 10 years away. Now it feels like it's maybe five years away. So definitely feel that kind of sense of urgency. I didn’t realize how big the problem really was until I started applying for funds and grant applications and things like that to support the research, and it turns out that it's the single most expensive disease in this country. So Alzheimer's disease is one form of dementia, and it represents about two thirds of all dementia and Alzheimer's and related dementias are the single most expensive disease in the country. It's one in five Medicare dollars. It's one in three people over 85 and one in nine people over 65, which just does not feel old to me. You know, like, you know, my parents are over 65 and so many others, you know, it's hard. If there's two of us on this call, you know, it's a pretty good chance one of us is going to have, you know, in the period of, you know, deteriorating brain damage and it's a really hard thing to deal with for both the individual, of course, but also their families and requires a lot of support. And I think one of the things that makes the disease so tough is that it doesn't kill you. You know, folks with Alzheimer's will live with it for five to 20 years and and during that time. You know, they're going to progressively lose their memory. And so, you know, people know about things like, well, forget the people around them and things like that. We also lose the ability to do anything on your own. You know, you lose the ability to eat on your own, get yourself dressed and things like that so that you get to the point where you need 24/7 support. And it's really hard. It's really hard work. And as the population is aging, we're now at a point where there are more people over 65 than under 65 for the first time in human history. And we just do not have good solutions. So, yeah, very eager to help my own family, but also that so many really need us and will just become an increasing challenge for our society on the same level as climate change and things like that, problems that we have to solve. Or they will just increasingly make it really hard for a society to do anything else. 

 

Pieter: It's such a big challenge, of course. Also, it’s a big open problem. I got to imagine you are at the time still an undergraduate student thinking about what you want to do with your life, and you take this as your mission to to improve. Our ability to care for people with dementia or ideally solve dementia, of course. How do you even get started? How do you say, “Well, this is the part of the puzzle that I can maybe make some progress on, and it's an important part of that puzzle.”?

 

George: One step at a time. You know, I think we in popular culture love to think of start ups. It's like, I don't know, you see Elon Musk. So it's like, Wow, what a great idea. Whatever else, how did you come up with that? And it's so at least for me, it was not like the sudden burst of insight or whatever. It was very iterative, going and talking to a lot of people exploring different ideas, thinking about what's the biggest risk to prevent this from working and kind of just taking it one day at a time and being patient and, you know, seeing when you're really getting that jaw dropping reaction from folks. And so when I was an undergrad, to go back to your question, both my parents are immigrants from South Africa and both are physicians and, you know, worked really hard to get to this country and find great fulfillment from, you know, doing good for people. So I think that's really kind of core DNA that I really wanted to, you know, do something that made the world better. And I thought I was going to go be a doctor. But then you fall in love with engineering, and so many of us have like, it's really fun and you see the opportunity to help so many more people, you know, instead of just helping a person right in front of me. I can build something and that work can help. So many others think so many engineers come to that conclusion. And so then apply it to grad school with really just that idea of somehow combine these new tools in the population I care a lot about and I don't know how, but it's got to be some, some way we can support. And so this what became say for you was actually we worked on several projects while at Berkeley exploring lots of different ideas. And then we were doing this work off campus and we saw the eyes really lighting up. This is something we can solve the problem today for folks, and yes, there’re all sorts of risks associated that we think we're going to be able to figure out as the company grows. But here's a way that we can both help right now with the kind of sense of urgency I feel about my own family. But also, oh crap. Once the system is in people's rooms, there's only going to be more and more that we can do from that, from the data and from the population to see a trajectory for. You know, not just solving really hard issues today, but actually really solving in a systematic way some of the challenges that we are already facing and will continue to face as a society. 

 

Pieter: Now let say somebody gets a SafelyYou system installed in their home. What does that mean for the life of a patient? What does it do for them? 

 

George: Yeah. So basically, what the SafelyYou product is today is that we put cameras in people's rooms if they choose to have them and then we detect if somebody has had a fall and then we only keep that video, which is really the key point. So the goal is to give a voice to somebody that can't necessarily advocate for themselves. So what we do is we detect if somebody is gone to the ground, we only keep that video. We do it with super high accuracy because we have the largest ever dataset that's ever been collected in this space. We make that video immediately available. Where the big challenge is that these folks go to the ground and they can't necessarily get back up on their own and they can't tell us what happened. So that means that we have a lot of what we call unwitnessed falls. We find somebody on the floor. They could have hit their head. If you asked them, Do you hit their head? They're going to say, yes, you know, either way. And so you really don't know what happened. And so that means we have kind of everything to keep this person safe and we don't know what happened. So we have to assume the worst. So that means often we're sending folks to the emergency room because they could have hit their head and we don't know. But it turns out that what we consistently see is about half the time the person actually went to the ground intentionally. You know, somebody got on the ground to pray, for instance, and they didn't fall at all. But they can't get back up on their own. They can't tell you why they were on the floor and they don't know if they hit their head or not. And so they're getting sent to the emergency room and then that just perpetuates in the emergency room. We don't know what happened, and so we have to make sure this person's safe. And so we're going to do a full workup. We don't know if they hit their head or hit their hips. We're going to scan every part of the body and if nothing comes back on that scans well, they could still have a latent injury that just hasn't shown up on the scans yet. So we're gonna hold him overnight for observation, and then we're gonna scan them again the next morning and kind of the best situations for the costs, you know, thousands or tens of thousands of dollars and in the worst situations were exposing folks to COVID that are hyper vulnerable. And then we're sending them back and exposing other people. And we heard some really horror stories that actually happened to my own family as well people sent to the emergency room during COVID. And families can't go in with their loved ones, right? They're restricting it. And so you're sending somebody with Alzheimer's disease into an emergency room and the family is sitting outside in the parking lot and they don't know if their loved one is having a brain bleed right now. It isn't even being seen like they could be sitting in the corner of the emergency room and nobody knows they're there and they're, you know, having a brain bleed and families are sitting there thinking about worst case scenarios in the parking lot, you know, 20 feet away from their loved ones. And then, you know, only find out two hours later they've actually been sent to another hospital because that's the one with the neuro center or whatever else. And you're just, yeah. So I think, you know, detect the fall. So we can, you know, respond right away and be able to see how it happened right away for a population that can't advocate for themselves. And then the third big piece is we provide expertise, so we have experts on our team. Often, occupational therapists review every video and work with the families or the staff to identify how this person is going to ground and what changes we can make to keep this person safer. So really, what we've built is kind of this room remote care model where you don't need to be there all the time and the expertise is there remotely. So wherever you are, you can be providing exceptional care for your loved one. 

 

Pieter: So there are these cameras that are set up in people's homes and then you have an AI system that is somehow monitoring this camera feeds. Can you see a bit more about how it's built? 

 

George: Yeah, happy to. So at a high level, we're using a deep convolutional neural networks. And our aim is to detect specific events of high importance. So really supervised learning techniques where the goal is to take, you know, a video in and have a confidence score out. So we want to take a video in and no, as the end result, how likely is it that this person is on the ground and particularly in our problem? You know, when you think about error modes, we think about precision and recall or, you know, sensitivity and specificity. And we care a lot about never missing a fault. And so really, what we're training is to be in this regime where we have almost always detect when a fall happens, even if it's at the expense of greater false alarms. And then a lot of our work from getting a very high, you know, highly sensitive algorithm was how do we reduce our false alarm rate so that we're not overburdening staff and things like that? There's a lot of challenges with alarm fatigue in the industry. So as of this year, we did an analysis at the end of Q2 of this year, and as of the end of Q2, we were at and we require our customers to report back to us. You know, we work with them through the whole, you know, human expertise side. We find out if we ever miss a fall so we can track pretty accurately how well we're actually doing. And as of the end of Q2, we were detecting ninety nine point three percent of the events. So missing seven and a thousand, which we take a lot of pride in and crazy situations, right? I think all of us, when you do deploy models, you start to be amazed at how well they actually working. Sometimes it's like we only saw this person's legs sticking out from behind the bed and we still got it feels pretty cool. And then in terms of the alarm rate we send out today, one false alarm every two years from each camera, which feels just insane. And that's where I am extremely happy with where our with our error rate is today. 

 

Pieter: That's amazing. One false alarm every two years. I mean, that means, you know, because there's always to talk about if you sent too many false alarms, people stop paying attention to the alarm once every two years. I don't think anybody will fail to pay attention. 

 

George: Yeah, that's right. A lot of hard work over a lot of them. Yeah. 

 

Pieter: Do you have a sense of, you know, how much data you collected in this process to build this kind of system? 

 

George: Yeah, I mean, hundreds and hundreds of terabytes at this point, it's hard to, I don't know, have kind of specific math cameras for a long time. What we required is that the cameras were recording all the time and which is a hard ask from people, so we had pretty low opt in rates and things like that until we could get up to the point where we had enough data that we could do things with high enough accuracy that we could start to have these of privacy protections in place for folks. And and so, yeah, we spent a really long time in that mode, and I think we could only do it, you know, by virtue of. The personal connection of, you know, hey, I was going to individual families and saying, Hey, I want to put a camera in your room and it's in a record video all the time. And I think if we do that, I can build something for my own mom. And you know, it might not help your loved one today, but it might help the people that come after you get folks coming up to you to get back. And so it took a long time. You know, we were on SBIR funded grants and things like that to run research studies with IAB approval and. I think if we were, there's big chasms to cross there and if we were in it just for the money, whatever else, there are much easier businesses to build. You know, when you recognize how sensitive that data is, I feel very fortunate to be where we are now. 

 

Pieter: Now we talked about the technology, but as I understand it, I mean, you really come from the mission that technology is just a means to an end. And as I understand it, you augment the A.I. system with human care. Can I say a bit about how that works together? 

 

George: Yeah, I think it's a really critical piece. Because the problems are really hard and the solutions today can't really provide complete solutions, right? So, you know, as an example, having the human expertise of not just having the video available, but helping folks know what to do about it is a big part of the secret sauce. I mean, not so secret sauce. Like being able to actually make meaningful change…if you think about, say, an assisted living provider. Unfortunately, detecting a fall and having their staff get there faster does not make them more money or save them money in any way. So there's no ROI for them. And so even if they might really want to do it because they know it'll make life better for folks. How do they pay for it? As we were running our research studies, one of the things we did was we had our cameras on without sending alerts out to the communities because we wanted to test how long are people actually on the ground normally without our system in place, and we did it all with IAB approval and families were consented and residents. It turns out that basically their standard of care is that they'll check on residents, usually every hour that are high risk, they'll come in and do rounds on them. Well, what we saw was that the average time someone was on the ground was 40 minutes. They had a fall, and that's the average. They can be on the ground for a lot longer than that. And it's really tough, but they want to be able to move forward with solutions and things like that, it's such a tough industry. They're on such thin margins and things already that it's really hard for them to move forward with things unless they're really seeing ROI from it. And it turns out that there is really meaningful relief from actually reducing the number of events. How do you have fewer falls? How do you provide better care for people? And in ways that take a meaningful amount of work off the staff and things like that. And so that piece of having the human component and not just throwing technology at them, but really make sure that the technology is successful. I think many technology companies are very resistant to having that much of a human component because they want very high margins and things like that. So, you know, how do you find ways to have good margins and be a venture backed business and all of that, but still have a complete solution where I think a lot of AI systems or technology systems in general expect the end user to kind of close the gap for them. But in underserved industries and the industry is on thin margins and things like that, they don't have the staff to really kind of close that gap. And so it's really up to you if you want to make sure it's successful and really help folks to go that extra mile with them. 

 

Pieter: Kind of curious with everything you described…you need some kind of right to be able to to work on this, of course, because you need to build, you know, pay people and so forth. And when you think about the ROI, it seems like there's a very big ROI from the perspective of what I think would often be the children of the patients because they would see a higher quality of care. Is there ever any interest from from the children of the patients that say, “Well, this is so much better for my parents and it cost more. That's fine. Or I'll save money from E.R. visits that we'd be doing otherwise”…things like that?

 

George: Yeah. And just one minor note…not to call you out, but we don't love to use the word patient unless it's like somebody who is in a hospital setting or things like that. And usually we say like ‘people living with Alzheimer's’ or things like that, we don't like to be defined by, you know, we're going to live with it for 20 years. I don't want to be a patient for 20 years. You know, I'm just George and I happen to have Alzheimer's. Yeah. So, yeah, but then at the same time, I think your question is a really savvy and good one. And I think one that all entrepreneurs put first and foremost is we want to do a lot of good. But the only way this gets out to as many people as we want to help, it has to be a financially sustainable rate. It has to make people more. If you're selling to enterprises, it has to make them more money, right? Or else they're not gonna be able to buy it in a scalable way. They might be able to do it as a “nice-to-have”, but for it to be a “must-have” and become the standard of care. It has to drive the businesses forward. And so you asked kind of specifically about the children. I think a lot of people have gotten stuck there. So a lot of companies going to support elder care, which, by the way, I wish there were more of there are not enough. We, as a society don't love thinking about elder care, but a lot of companies that have approached elder care have unfortunately kind of floundered and failed. And so, you know, a lot of VCs don't love this space, and they've had, you know, stories that they thought would be successful and just didn't find the market traction that they hoped for. A lot of companies have gone and specifically tried to market to the adult child, and that's something that they can pay for for their families. The challenge that a lot of these companies have seen is that the individual has to also really want it, or you're going to buy like one of those pendants or Apple Watch or whatever and. You buy it for your loved one and then they go throw it in a drawer because they don't like being spied on or whatever else. So one of our things that we take great pride in is that and so most of our customers are living in facilities so either assisted living or skilled nursing facilities. And in the rooms where it's offered, we get 90 percent of them. We track it quarter over quarter, how many people are opting in for our system. And so that's, you know, the residents and families that are explicitly choosing, we want this system because it makes our own lives better. And so it's a really delicate balance to play. And I think any probably any company will tell you, but certainly any health care company will tell you is that the big challenge with any health care company is that the person who benefits from it, which in our case is that resident or the individual Alzheimer's is different from the person who's paying for it, which might be the business or the family is different from the person who making a decision about it, which might be like the senior executive within the company. Or it might be that the health insurance is paying for and there's a totally different entity that's paying for it, whatever else. And one of the things I found really hard and trying to help folks with Alzheimer's is that every single one of these stakeholders has to say yes and enthusiastically yes. And if any one of them says no and it's just not going to happen, like if the staff don't like using it because it makes their jobs harder or whatever else, then it's just not going to get used, even if it does a lot of good for the individual even to make the business more money. And I guess particularly, that was really hard in the early days so I really focused on how do I help my own, I really wanted to help the individual themselves. And then you start to understand it's really hard to help somebody who can't necessarily learn to use technology and whose needs change over time. So the technology that works for them today, two years from now, doesn't actually help them anymore. And people have experimented with things like Google Glass to like, Can I have whatever? And if so, then you start thinking about the ecosystem and how do I support their caregivers or things like that? 

 

Pieter: one step at a time. Now, one really interesting statistic to me is that you've already achieved 40 percent less falls thanks to the SafelyYou system. But then when I naively think about it, I just think like, OK, a camera is looking at a person, how does a camera looki at a person, maybe with an eye behind him, but still just looking? How does that prevent falls? Can you expand a bit on that? 

 

George: Yeah. So as you were talking about earlier, it's 100 percent that human expertise side of it. And so I think kind of a big misconception around falls is we think of every fall as being this very dramatic incident. It turns out that only about five percent of falls result in an injury of any kind. Only about one percent result in a severe injury, like a fracture head injury. So the best practice today outside of our system is basically anytime somebody has a fall, do a root cause analysis out of this. First of all, what kinds of things can we change or be revisiting their medications should they be on physical therapy? Should we have a nightlight in their room so they can see the bathroom? And so for folks that are cognitively healthy, they can remember what happened and tell you what they need and know if you or I, you know, stubbed our toe going to the bathroom at night, we would remember that and like, hopefully do something different the next time. These folks unfortunately have a very high rate of repeat falls because they don't necessarily learn from their previous falls and falling the same way over and over until they have that fall with injury. And so really, what we've been able to is that same root cause analysis process. But for somebody that can't advocate for themselves, how do we give them a voice where they would want to be able to tell us how they felt? They wouldn't want us to go to see if they were, you know, changing in their room or whatever else. So now we can see when they had a fall and how it happened, and then we can provide expertise where our staff are helping with, you know, thousands of dollars every month today and identify, Oh, this person, you know, we can see that they're getting out of bed on their left side and they're forgetting that they have muscle weakness on their left side. So they're going to the ground when they're trying to get out of bed. If we just push the bed up against the wall and they have to get out of bed on the right side, and we can kind of design environments where we can kind of learn for them and you're not going to prevent every fall, but you're going to, you know, understand it's kind of like a debugging tool for falls, really. At the end of the day, how is this person falling? What kinds of things can we put in place to try to keep it from happening again? 

 

Pieter: That's really interesting. Now, the technologists may think that at some point that could even be, you know, an AI system, too that has seen from many examples in the past what has been done to improve the care and then say, Oh, wow, this room, I can already tell you ahead of time. Here are some weaknesses for this fall. I know how to prevent that. Have you thought about that at all yet? 

 

George: Without a doubt, I think we're just scratching the surface I hope. So kind of where we are stage wise as a company, in May, we did our Series A, which is 20 million. And in September, we did our Series B just a few months later, which was an additional 40 million on top. And obviously, that only comes if we're, you know, growing really quickly as a business and helping a lot of folks. And so really, where we are at stage-wise is we've proven that there's a scalable solution here, but we are just scratching the surface. And so we're applying a lot of those funds to work on exactly the types of things you're describing. And that's just one of several things that people don't even aren't even thinking about today because they don't seem technically feasible. You know, like fall detection is so clearly just the beginning of, you know, once we can get a camera in somebody's room and start detecting different types of events, and there are so many ways that we can provide more and more support. And yeah, certainly we're collecting the way that our human experts work today. Their workflows are explicitly designed so that that data can be used to train algorithms with their expertise as label training data to provide more and more support, and for us, that's a big part of the mission. So in our mission statement, we want to push the boundaries of dementia care, make care better than it ever could be without us. But we also want to make it accessible to everybody. And so we see those things as how do we make it accessible to everybody and really bring down the cost so that, you know, wherever you are? And because unfortunately, a lot of folks are on Medicaid and low income and don't have the support that we need today, and how do we make it accessible to everybody while using the AI to take over more and more of the human expertise is a big piece of it. 

 

Pieter: Now, George, you said Series A and in May, it was Series B. To raise in the same year seems to indicate that there's a lot of momentum and that your technology is being adopted in many, many places. Can you say a little bit about that? What's going on in terms of growth? 

 

George: Yeah. There was a bit of that start-up rollercoaster of when COVID hit. We didn't know if we still had a business, right? So March 2020, we actually had a term sheet for the Series A. That was like weeks before that. And when COVID hit our terms, it got pulled because we didn't know if there still would be a business here in a post-COVID world. And we're still not in a post-COVID world. Hopefully we get there someday. We went in a year from not knowing if this business even make sense anymore to some really rapid growth. And you know, a lot of people might think, Oh, well, I'll put this a remote care solution like clearly that works in the COVID world to be able to support people without necessarily needing someone in the room and whatever else. And what we saw was basically our sales went flat in Q2 because folks were just focused on infection control and the needs there. I mean, in Q2 of 2020, and so we ended up doing about 4x growth that year. We would have straight up down about 8x growth if you kind of annualize that chunk of the year out. And so basically what we saw as the solution was driving a lot of good for folks and then being able to keep folks out of the emergency room in a COVID world, just that all the more value we don't to get them exposed to COVID, but also the industry. We served their budgets now like they're in recession, right? Their revenues are down, their costs are up. They don't know if their businesses still make sense or what the long term trajectory looks like. And there is still a lot of questions, and we basically set, you know, in 2020, these are company goals. We want to show that we can still be close to customers. We want to show that we can still grow those customers. And if we hit the key goals, then there's a business here that still makes a lot of sense and a lot of, you know, good we can do for folks. We came out of 2020 hitting that same hockey stick we had hit the year before, and so a big reason why we could raise as much as we did is that. You know, I think very few companies have to show the hockey stick twice, but you know, seeing it twice gives you a lot of confidence in it that it's not just like a one time thing. Even through all sorts of challenges, this team is going to make this happen kind of no matter what. And so really thrilled with the growth. We got to sign our first eight figure contract this year and things like that that, yeah, thank you. One step at a time. But yeah, it feels really good to be at a point where just so clearly so much of what we do makes sense on a lot of levels for a lot of different stakeholders. And now it's just how can we when we do more and more and more and more folks? So I think my big ask, is anybody listening, we are hiring across the company. So if you want to make the world better, we want to hire you. You know, if you're a top caliber type of person and we hold a pretty high bar, but everyone we look for, we want them to be people that are interested in making a big positive impact, particularly folks personally affected by Alzheimer's. And so if you're any type of role, you know, if you're interested in why you're interested in sales and marketing, you're interested in just helping in whatever way. Maybe you're just a super type-A person and you'd be great on the outside to make sure that we're not dropping any balls. We would love to meet you and I appreciate the opportunity. 

 

Pieter: Well, great opportunity here to make a big difference in the world now.. So I'm curious, with all your experience in the space, what advice would you give someone that had a family member diagnosed with Alzheimer's? 

 

George: That's a great question. You know, this is only my personal perspective. There are many others who can probably share better insights than me. A book that was really helpful for me was a book called, “The 36 Hour Day”, and I also spent a lot of time sitting in on support groups just to understand what other folks are going through and what the challenges are and things like that. You know, the emotional part of it is really hard. And so basically, don't blame yourself too much. I guess the big advice that's given is come to their reality, you know, don't try to correct them or, you know, they might wake up in the morning and think that they're supposed to be in the house they lived in 50 years ago and then go running down the street because they're trying to find their house. And you can obviously imagine how scary that must be for them. A lot of the advice that's given is to come to their reality and don't try to correct them and whatever else. And that obviously takes a lot of love and patience and emotional challenges. You're progressively losing somebody, right? Like you're seeing it, you know, has highs and lows of some days are better than others. But over time progressively losing this person more and more and you're grieving for that. But you're also having to do more and more for them. And it can be very isolating where you might not want to go out and eat at restaurants anymore because, you know, they might have outbursts or whatever else. And so you're also feeling frustrated and angry at them for and you try to direct it towards the disease and it's not them. But if you feel frustrated with losing your own ability that it feels like you can't do anything but care for this person. And if you have a family support network that can do so much for the means to have, you know, people that can come in the home and help you, you're dealing with losing this person, then you're getting angry with them and then you're feeling guilty about that, right? Like that you don't. You feel angry, but you know you shouldn't or whatever. And I don't know. So I think the advice that's given is try to come to their reality. My advice would be like, don't be too hard on yourself, just like, know that it's really hard and you're doing your best and they're lucky to have you and, you know, lucky to have the family around them. And there's a lot of other people struggling and support groups can be really helpful. You know, get a chance to both empathize, but also laugh at the crazy, crazy stories that other people, you know, have as well, and you know, you can't do anything but laugh at some of those times. So you know, it's, you know, we do everything we can for the people we love, and it's a terrible disease that hopefully, you know, I want to help people today. So I wasn't interested in the drug development side, but hopefully on the drug side, we will have therapies that come out and can help. But. You know, we have such a lack of understanding of how the brain really works. I was listening to, I think I think it was your podcast, the same thing I was listening to. A little of you said something like it felt way too hard to understand how the brain works and easier for us to build our own thing and try to replicate. It's so true. We have such a lack of understanding of how the brain really works that in Alzheimer's case, it's really hard for us to build effective therapies, and we really don't know what the root cause is. We see plaques form in the brain, and it's not known if those are symptoms and there's an underlying cause and we can give things like a knee transplant. We can't give up on brain transplants. And maybe the end result will be that we find some way to upload our consciousness to a server and they can download it into a clone or whatever. I don't know, but that it feels like it's still a long path from a therapy standpoint. So for now, I think we just do everything we can to try to help each other. 

 

Pieter: Kind of curious, when you look at the future of care for for Alzheimer's, right? What are some of the things that you're seeing? In the near future and then the further future as opportunities where we can actually try to make progress and help the situation and hopefully solve it? 

 

George: Yeah. So in terms of solving it, Alzheimer's dementia is one form of cognitive impairment, and this is even more folks with cognitive impairment, right, that have been in car accidents or, you know, veterans that have traumatic brain injury. So I don't really think in terms of like, how do we solve cognitive impairment as much as how do we just keep chipping away at making it easier and easier for folks? And I think that there's a lot of opportunity there from a kind of age and technology standpoint. I mean, some of the work you do in robotics, we'll see those things be deployed in places like manufacturing and places where, you know, it's not such a vulnerable population. And then as we see the successes with the ability to grasp well and know when we're grasping too hard or too lightly as we see those successes in safer environments will come to places where, like elder care, there was a robot that could help folks go to the bathroom independently. That would be amazing, and it would take so much off like I think I see a world where, you know, loved ones can increasingly do the more fun, pleasant work and less of the butt wiping and stuff like that. And so I think that there is that world seems like clearly feasible and it's just a matter of time and engineering effort and they're going out the business models and proving successes in certain industries and then bringing them to others. And so I think there's a lot of opportunity there. And then I think just through, you know, Zoom and some of the remote connectivity technologies. I think we'll figure out ways to also support on the companionship side where families, if you're like a husband and wife and your husband has Alzheimer's and you're living alone with them, you can get just totally worn down. And and to being able to be, you know, a son or a nephew and able to still spend meaningful time with somebody of Alzheimer's to help kind of cover a shift and just be with them in a meaningful way remotely. I think those things will keep making progress. And maybe there's some cool stuff from the VR standpoint where, yeah, you know, it's surprising actually that folks with Alzheimer's, they can use stuff like that. And it's like not to go on a little rant here, but also a lot of times are really good at covering that they have Alzheimer's, and they're really good at still having a normal conversation and within one conversation, you don't necessarily realize. And same thing with like, you know, we'll get on video calls and people will just roll with it. You know, they might like talking to a computer because I don't want to admit that I don't actually know that this is a normal thing and because people are embarrassed or whatever, and so they just roll with it. And so I think that there are a lot of times they do, and obviously it depends on the stage and whatever else. I think there's opportunities from the VR standpoint, you know, people that are isolated to be able to bring connections to them. And there's some very cool companies that are doing cool things there already. And hopefully we see more and more of that.

Pieter: Related to that, Alzheimer's affects memory largely, right? And so I wonder to which extent you see any future in terms of somehow using the camera systems in a way that gives people a summary of their day? Or some kind of intelligent augmentation of, you know, when somebody is with them or is going to visit them, the kind of intelligence augmentation for the things that maybe they have a hard time retreating from their memory? If a camera was with them at all times, maybe they can retrieve that part of the memory and somehow re display it to help people kind of connect things?

 

George: Yeah. So I think those are really hard problems. So I think, you know, part of our vision of the company is that we want to be the perception for these folks, right? And to do that well is a really hard problem and takes a lot of time. And so we're kind of not going away. You know, one use case at a time, but that very much is a big part of the vision. The challenge that you see is that, you know. So, folks, you see if you Google research papers around like using Google Glass to support you with Alzheimer's, you'll see lots of people do things like that, but it ends up in the research setting sith problems that like, Oh, I can, you know, so kind of low tech solutions, people will use this. I'll go put sticky notes on the cabinets. Right? So instead of needing to go open up each cabinet, I have like, Oh, this one is clearly labeled that it has the serial in it. This one has balls in it. And so even if I don't remember, I can pretty clearly see. But then what you see is that the as time passes, those solutions start popping. And so what you really want is a technology that can also grow with individual, so recognize that their needs are changing and then still be able to be like a virtual assistant. And I think. There's different opportunities for that, but it's so it feels to be science fiction. And it feels in the sense that like there are toy problems that can make it look really good and flashy or whatever, but there are so many needs when you think about how hard of a problem that really is and you know, people can get frustrated easily and things like that. So setting the appropriate expectations about what works and what doesn't. But the UK does a grand challenge every year, and I believe their grand challenge is specifically on virtual assistance for people with Alzheimer's that can use machine learning to grow with them over time. So definitely people investing a lot of money in it, I think Bill Gates might be part of that. I'm not sure. I know he invested a lot of money on Alzheimer's solutions, and so I definitely think we'll get there. But I think it comes with how do you collect that data at scale to know what it's the meaningful stuff to be supporting with? And I think you start with specific use cases that show a lot of value around them and then you proliferate out from there. 

 

Pieter: Now if we kind of zoom out for a moment and think about the future of health care more generally, I'm curious if you have some thoughts on that. The combination of A.I. and health care beyond Alzheimer's and dementia. Just kind of big picture…any thoughts on where you see other opportunities for people to make big contributions and improve lives? 

 

George: Yeah, I mean, I'm sure you've spent a lot of time thinking about this as well so I'd love to actually hear your thoughts too. I guess in two broad buckets, I see the opportunity to push the standard of care to do things that a human would not be able to do. And then I see the opportunity to just reduce workload on humans. And so, you know, for I think radiology is a classic example of like there's a lot of X-rays and an AI algorithm can look at and doesn't actually need a human to look at to understand. And then there's an area in the middle that, hey, we're not exactly sure. Let's hand it off to a human expert and have them figure it out. And then if you have to work, what’s that look like that over time? You get better and better, and that decision boundary gets more and more clear. Obviously, you still need to balance it with your training on past data. And so there's a shift in your underlying data and whatever else you know, you need to be aware of those things. But I think we definitely see as particularly with an aging population as there are fewer and fewer young people available to care for the older folks, things that just let people be more efficient and only respond. One of the statistics I've heard is that in 2015, there were seven potential caregivers for every person over 80. By 2030, there will only be four. So that's not a long time. And yeah, we just have so few relatively young people to support all the older folks. So how do we make them more efficient? And then in time, how do you actually do a superhuman level of tasks where a human actually can't tell in the air can to a better degree or whatever else? I don't know. I'd love to hear your thoughts, how you think about these things. 

 

Pieter: Well, there are two things that come to my mind. One is it seems that, hopefully, AI could be really helpful in diagnostics. I mean, kind of what you're doing in some sense is diagnostic. Diagnosing what happened that led to the fall, right? And then from there finally finding an improvement in the process and so forth. But I'm really hopeful it could be really helpful and in better diagnostics also kind of the chemical level of any kind of blood samples, any kind of diagnosis that's, you know. Physical in some sense, and I'm also I'm very hopeful about the new efforts towards drug discovery, where instead of being all physical trial and error experimentation, where simulation and machine learning can start doing a lot more of the heavy lifting of the discovery process. Obviously, this is mostly in the future and it has to be proven out, but I'm really hopeful there might be some things there. I'm also really hopeful in the robotic side, which you already alluded to, that I think robots should be able to help out with a lot of the Kona less pleasant chores that need to happen. And I mean, one of my personal dreams remains that, you know, robots could help people to live independently much longer once it becomes too tiring to, let's say, cook every day, too tiring to clean up every day. Well, that doesn't. What if a robot could do that? Then maybe you could keep living at home. You could keep hosting your children, your grandchildren at the house you've always lived in because maintaining upkeep of the house is and just the day to day activities are not a burden they're taking care of for you, as you can keep your your place that you've had for so long and independence, which I mean, some statistics I've seen in the past have shown that once people feel they lost their independence, it often leads to kind of mentally a very big downward spiral. 

 

George: Yeah, it's very awkward. And I think funny enough, a lot of the services that have come out that enable convenience for folks. Like how San Francisco is effectively assisted living for like 20 year olds where you can get right, whatever you want, you can get food delivered, whatever. And so a lot of these things that have come from a convenience standpoint, Alexa, for instance, are really meaningful for the people that aren't getting them for convenience. That's actually filling a need for them, right? So Alexa, being able to change channels for, you know, folks with disabilities that may not otherwise be able to interact with the remote easily or whatever else and or just really hard for them to get up and go get the remote. And so it's expanding access and things like that. I think we're going to see, you know, when I hear you talk about like meal prep, for instance, it's like, Oh, yeah, that's definitely going to start with like, not explicitly designed for older adults. It's going to start with people who just don't want to cook and are willing to pay up for something for convenience. And then costs will come down and things will get more and more seamless. And then it's going to do a lot of good for the population that it's not just about convenience, it's about independence and being able to live a meaningful life. I love your point there. 

 

Pieter: It's another direction. Actually, I have a question for you, George, but different from what we're most talked about so far. I'm curious. Now your company. It's doing well, it's helping people in the world. You have multiple, you know, successful fundraising rounds and the system is being adopted more and more widely. And so I can't imagine that you also start getting questions from budding entrepreneurs for advice. What is some advice, some lessons learned for people who want to start their own companies? 

 

George: Yeah, I'm sure you get this as well. So again, I'm gonna want to hear your answer to this after. And I think mine is probably like we should work on something that you're willing to work on, even when it's really hard. You know, we're going something that you care a lot about where you're not going to push through some of the wall. I mean, I think there's so many if you Google Camera fall detection, you'll see papers going back 30 years, right? But you never saw a company really do it at scale and do it well. And there's a lot of challenges that are going to come up. And it's not obviously just the technology. It's going to be all sorts of, you know, how do you figure out what go to market motion looks like and what's your growth model and. And, you know, there was a time when I was there only on site support, so we were running our first studies to prove efficacy and had no money, and I would spending every Sunday driving all over the Bay Area and to Sacramento to just fix cameras that were down. And I remember being just so exhausted that I pulled into a community at 7 AM and just I cried in my car. I was so tired. Yeah. And you're not going to if you're in it just for the money. There are times where you’ll be like, “I feel great. And there's gonna be times you're like, Forget this.” There are easier ways to make money. And I kind of feel like we glorify entrepreneurs a lot in our culture, but I feel like it almost has to be. I don't know. Maybe it's to work on a type of company we built, but it almost has to be a calling where it's like I really could not have done anything else, like I needed to do this. And if I was going and working at a tech company or whatever on something else, I just would not. I would have been bored and not found fulfillment. And this is something that in my bones I needed to at least try. And so say, like, recognize how hard that can be and that I think the big successes come from people that are going to push through whatever wall gets put up in front of them. And then probably, you know, was just talking something else about this. I see your fundraising and hiring people and things like that. Don't be greedy. Don't get caught up in like trying to keep the biggest piece of the pie. Focus on just making it really big. Focus on rewarding the people around you and giving like meaningful amounts of equity to your employees and your co-founders. And, you know, don't get greedy with your fundraisers. Raise more than you think you'll need because guess what up it's going to hit and they'll be very happy, you know, whatever. So if you think first and foremost about the long term success and what is the best thing for the business and you know, everything else will come from there, if you really are very committed to this thing is my baby and I would feel, you know, just upset. You know, I guess I don't have children, so I probably can't make that statement. But I feel like, you know, the success of the company is my success. And you know, everything you know, the company comes first and foremost and you know, I've got my own pay three times when I was need it right before I knew if we had salary for other people and you should be, you know, very ready to. It's one thing to say, you know, I think one of my surprises is that it takes a long time, you know, everything takes longer than you think it will. And so you might feel like that may be true for you right now, but think about where you'll be seven years from now. And if you'll be wanting to have kids at that time and have other other people that you want to put first. And so. I think, you know, that's my personal experience from being a starting a company with no name first-time founder in a space, it's crazy hard and whatever else and I don't know what. What are your thoughts there? 

 

Pieter: Well, the first thing on my mind when you said you were going to ask the question back to me was actually the first thing you said, which is building a company takes a lot of work for a very long time. And so, I mean, certainly, you know, periodically there is an almost overnight success, meaning it only takes a few years to have clear success. But the norm is that it takes seven to 10 years to emerge as a truly successful company. Right? And that's a long time. And so, yeah, really wanting to build what you're setting out to build is key or you're just going to, you know, start doing something else, especially, you know, there's so many job opportunities out there that you know that something else could be much lighter on. You could be, you know, paying you better. I mean, a very successful company ultimately will likely pay you more than whatever you could have done by going work somewhere else, but even with a medium successful company, often, you know, if you climb the corporate ladder at a big company, you might even match it. If you cleverly climb the ladder and do well, so I think you really need to want to build what you're setting out to build.

 

George: Yeah, don't get me wrong. Definitely. Especially in the early days, how those nagging thoughts of like, oh, friends from Berkeley going and getting like crazy salaries at tech companies and whatever. And I think that's okay, right? Like, I think be honest with yourself like that is definitely OK to just like, do work. And it definitely can be, you know, fulfilling make a good salary and, you know, prioritize things in your life that you care about, like, you know, your significant other and time for travel and things like that. Because building the company is a very high risk, high reward endeavor like most companies fail and you don't necessarily hear about them. And I think it's OK to be, you know, honest with yourself about, you know, where your priorities are in life. 

 

Pieter: Yeah. And then the other thing that would come to my mind, which you kind of alluded to also is to just have a really good team that you're going to spend a lot of time together and you want to just have a great team where it's really fun to to work together. And also, you know that you're all, you know, super driven and good at what you're supposed to do at the company, so you have a real chance of succeeding. 

 

George: Yeah, I couldn't agree more. And I think people where we found a lot of success, I think about our co-founder or our chief operating officer shortly or our chief strategy officer. And I think where I found a lot of success is people that share the same values. But actually think quite differently. So they are working on the same something but how we approach it in the perspectives we come from and what we prioritize. It's quite different and fills a gap for me, so if I'm a technology person, I really like to think about products, for instance, having people that. The why is the same, so to your point, we really like spending time around each other, we're in it for the same reason and we know we're going to push super hard and have each other's backs and all of that, but they're going to think about like customers and sales and go to market and things that are so important for our business to be successful. Or even you can both be thinking about product. You know, it's just having a conversation with our COO this morning about product opportunities and we just approach it in such a different way that both of them, both paths are clearly extremely valuable, but I'm thinking really from an A.I. standpoint. And she's really thinking from like a workflow standpoint, and both of those are so important. And so, yeah, finding people that I mean, so the value of diversity and inclusion and all of that, I think all the statistics are known, but I cannot really see that and working on teams like that or having people that are there for the same reason and share values, but think totally differently from you. You've uncovered so many blind spots and opportunities you wouldn't have thought about or prioritized. And yeah, a fun ride so far. 

 

Pieter: I love that lesson there, George, thank you. So great to have you on. I really enjoyed this conversation. 

 

George: Thanks for having me on.