OpenAI VP: How AI Changed My Family’s Fight Against Cancer | Ashley Alexander

There was a period around the pandemic where both of my parents each got diagnosed with two cancers, each within a week of each other. My mom had breast cancer three times, which is pretty crazy. I felt like I had an ability to potentially change the outcome based on how we steered the treatment path, and I didn't feel like I had full information. As the doctor is sharing things with me, a lot of time I'm using it to sort of real time, like, what does this mean for Dad? Like, what questions should I be asking right now? I feel way more empowered now to do that job for my family and for myself than I did without Chat. All right, Ashley, I'm really excited to have you here. I think this interview is really special to me because we both care for parents through cancer, and I really want to understand—there's a lot of hype out there—I really want to understand how close we are, how close AI is to helping patients, caregivers, and also physicians, clinicians. So why don't we start with just, like, can you talk about your journey caring for your parents and how it led you to OpenAI? Yeah, absolutely. Well, thank you so much for the time to speak about this and for your interest in this space. These are the conversations that I hope many of us will have more and more so that we can accelerate the positive benefit that AI can bring to all of us. It's like you know me. This is actually—some people don't realize this is actually the origin story of how I came to be here doing this work, and it dates back to over a decade ago when I was a sophomore in undergraduate studies. My dad was diagnosed with non-Hodgkin's lymphoma. He's in remission. Actually, since then, both of my parents have had two different cancers each, diagnosed closely with each other. Everyone's in remission and okay now. But through that experience, I really played this role of caregiver, advocate, companion. I go with them to all of their appointments. I'm taking notes. I'm chiming in with questions on their behalf. A lot of the time, especially in that very first time when my dad got non-Hodgkin's lymphoma, we were faced with a lot of decisions where, you know, there isn't a clear mandate: this is the treatment path. A lot of time—and for me, the first time seeing this in the system, the fact that we were kind of overwhelmed with so much information and yet still, you know, obviously in partnership with the professionals, but still, like, they would—it was like, do you want to do this surgery or this radiation treatment or this—you know, there's options often, and that can feel really overwhelming when, you know, I'm not the expert. I didn't go to med school. I don't have a background in pulmonology. And not only that, but— But all the information was so fragmented. You know, for my dad, the lungs and the heart, they're connected, but the pulmonology and the cardiology departments, even at some of the best institutions, which, you know, we were lucky enough to send him to, they don't. They don't. Yeah, they don't. They're not structurally set up to have that kind of. And so I'm the one piecing together that information on behalf of him. It's just a lot of burden, and it's a really high anxiety time. So anyway, that just all gave me this passion to say, you know, there are a lot of, like, smart and wonderful individuals in the system. The system is failing us. The incentives are misaligned. There's a ton of information asymmetry. I don't feel like I'm in the driver's seat of something that's the most important to me, which is my father's life. And we just, like, deserve better as humanity, is how I feel. So that was, you know, a long time ago, and I took a bit of a windy road. I studied healthcare management as one of my concentrations in undergraduate studies. I worked at Bain & Company for a bit. I did a bunch of healthcare cases there. Ultimately went into product because I'm an engineer at heart, and I like building things, and I wanted to build this skill set. But I always thought, even at that moment, I'm taking a quick detour. It ended up being a little bit longer than I thought, but a detour from my personal mission to one day find a way to transpose the skill set back onto this problem that I think is probably the most important work I'll do in my life. And everything changed. You know, I was always looking for an opportunity to transpose that skill set back, and everything changed when AI, you know, really when ChatGPT launched. I felt for the first time, you know, this difference and shift and the kind of information empowerment people can have in their own health, and it really changed everything for me. Yeah, I had a one-year-old and a three-year-old, actually. So, like, I had the one-year-old before ChatGPT launched, and I had the three-year-old after. And so I had my own experience with the health system of having babies, one with the power of Chat, one without, and it just totally changed the game for me. So anyway, I felt for the first time my skill set can be something that really impacts this problem at scale, and the world, and we as humanity, like, need this and deserve this, and I want to do everything I can to accelerate the timeline. That's awesome. I think I'll just share my personal story too. Like, we were talking about my mom had breast cancer three times, which is— Which is pretty crazy. The first couple of times, like I was still younger, and this time, like now I'm actually unemployed, so I have more time to take care of her. It requires that, doesn't it? I mean, it's a full-time job. It's a full-time job. You have to, like, so what I've been doing is, like, like you said, basically PMing the shit out of this thing, PMing the shit out of this thing, and, like, you know, using ChatGPT, using project, organizing all her information. Like, the problem is, like, a lot of the healthcare stuff, like, you know, doctors save lives and all the power to them, but the system is, like, you gotta, like, get fax numbers and, like, make phone calls, and, like, it's hard to even get their email address. So, like, that's what I've been struggling with, just, like, calling them every day and kind of being a pest and be like, hey, when can we get a diagnosis? When can we get the biopsy? and so on and so forth. Absolutely. Yeah, you have to advocate. That word advocate is a real thing, and I think about all the people who don't have a you or me. They're kind of fighting on their behalf, and how much, you know, they probably feel lost and scared in those moments, because we felt that way. And, you know, I come from, like, a middle-class family. We had insurance. You know, we're in the Bay Area where there's a lot of great access to great care. And even still, I felt scared and not empowered. And so imagine how it feels for the single mom in the middle of America who's trying to make ends meet. Maybe she's working multiple jobs and, you know, maybe doesn't even have full insurance. I don't know. I can't even imagine how that must feel for that kind of person to go through a big health crisis. So I think what you're doing for your mom is incredible, and I hope that we can build products that can give that kind of support. Maybe it can't fully replicate how amazing you are, but for someone whose alternative is nothing, we deserve as humans more support than that when we're going through a health crisis. This episode is brought to you by Granola. If you're in back-to-back meetings, you know how much work it is to take notes live and clean them up afterwards. That's why I love Granola, the best AI meeting notes app in the market. Here's how I use it. Granola automatically takes notes during a meeting, and I can add my own notes too. After the meeting ends, I use a Granola recipe to extract clear takeaways and next steps in the exact format that I want. Then I can just share the notes directly in Slack with my colleagues, or even get Granola to share the notes automatically. Honestly, of all the AI apps that I use, Granola is the one that saves me the most time. Try it now at granola.ai/peter and use the code PETER to sign up and get three months free. That's granola.ai/peter. Now back to our episode. Yeah, and can we just talk about, like, how I'm using ChatGPT real quick? Yes, tell me. Yes, I'd love to hear your setup. Yeah, so I basically gifted them, like, a ChatGPT subscription. And, you know, like, a lot of times the people who are sick are older; they don't actually know how to use all this stuff. But we basically just created a shared project, and we had, like, a Google Drive or something, and I started uploading all her medical records there. Once it has that context, it's actually able to provide, like, really good advice for better or worse, you know? And actually, like, a lot of it, like, I thought I'd be using it to do research on a disease, but a lot of it is just like, okay, who should I call next? Or, like, you know, who should I try to coordinate with next? And also a lot of it, honestly, is, like, personal therapy, like, because, like, you know. Like, I think we're both a little bit of control freaks, you know. Yes. But, like, you're right for me at least. Yeah, yeah. So, like, I feel like I almost feel like I can talk to my AI more than, honestly, like, my mom or, you know, because sometimes I don't want to tell her, like, all the resources and stuff. Sure. So, like, kind of getting advice from the AI and, like, kind of, like, figuring stuff out is super helpful. Oh, that's a great point, and I feel the same way. In large part, you know, the average doctor appointment in the United States is something like 15 minutes, and my experience has been even shorter than that in a lot of these ones. Again, not their fault. It's the system setup that makes it that way. And what I found is really helpful about AI is a lot of the time I need more time to sort of digest what's going on, to ask the follow-up questions, and I just don't get that time with a doctor. And so what I found is I'm able to do that in between the appointments so that I'm going in with much more focused questions that, you know, only the doctor can get to because they're so personalized to my dad's case. Yeah. Versus these kind of, you know, there was a new therapy that came out recently for his interstitial lung disease, and I kind of vaguely knew that the clinical trials were happening, and instead of using some of those precious 15 minutes with the doctor to ask, What's the latest? Can he try it? Whatever, in advance of the upcoming appointment, I was able to clear all of that with Chat so that I could go in and say, Okay, this is what I've learned. Do you agree with this sort of, like, take on where the clinical trials are at? Do you think we should try it with my dad? Should we not? And it just got to much more action planning. Yeah. Because I am a control freak too, and that just helps me feel much more prepared. And then when we're making those decisions, I feel much more confident in what we're doing. Yeah. And I understand what the doctor is telling me, as opposed to, like, being overwhelmed with this deluge of information that I just can't digest all in one 15-minute slot. Yeah, yeah. I've asked my parents to include me in all the doctor conversations. They're up in Canada, but, like, down in Maine. But actually, like, now that you talk to doctors and clinicians, I'm sure, like, are doctors kind of annoyed by, like, people like us coming with all these ideas, or can they deal with it? You know, so I talk to a lot of... A lot of clinicians, because we also build products in addition to building consumer products to support people in their health, like you and I are using to support our parents and ourselves, we also build products for clinicians. So I talk to them a lot. We also have many physicians that work with us. We have hundreds of clinicians that are red teaming and evaluating every major step of training. Has health context involved? This is something we take very seriously in terms of making sure the models live up to the sort of responsibility and safety standards that we do. Anyway, all that to say, I've talked to a lot of clinicians. Yeah, yeah. I definitely think there's both opinions. I hear both opinions. I would say, though, on balance, the majority tend to like that their patients can come in much more educated. And I don't want to sugarcoat it. There's definitely some clinicians who will call out the edge cases or the sort of like, it's sort of as reminiscent of the Dr. Google days where some clinicians don't like that. But I would say the vast majority of them have this reaction of, Wow, is this actually really a helpful tool? My patients are coming in more informed. We're able to make better use of the time. Because most clinicians didn't get into the profession to have to just fill out a bunch of charts and paperwork and make people feel overwhelmed. They don't like that either. Most of them came in because they really want to help people and they want to heal people. And a huge part of the healing process is understanding what's happening to you and understanding your options and being able to move through it together. Like so much of what a doctor does is so much more than just the actual process of prescribing the charting, the whatever. So yeah, most of them, I think, from what I hear, like when their patients come in more educated and ready to have the conversation. I guess just also, as a caregiver, you can also just talk to it the right way, right? Just kind of like cross-functional alignment, right? You shouldn't come in with like, Hey, you should go do this. You should be like, Hey, you're the expert, but I did some research and here's what I think I should do. That is such a good point, and maybe a good kind of PSA for all the people out there of, this is not a replacement for a doctor. It is not for diagnosis and treatment, but it can really equip you with information. And so you're right, I think the way to use this tool in a way that really actually strengthens your relationship with your doctor, if you can go in saying exactly that, like you coming to you as the expert, but I've done some homework. So that I'm coming in prepared with actually some pointed questions, so that I'm making better use of this limited time that we have together. I think doctors love that. Yeah. Yeah, yeah. I've only heard doctors use, like, you know, Dox or whatever, because, like, you're right, like a lot is riding on that 15-minute conversation. Yes. And if they had, like, these AI tools to help them process their emails or something, maybe they can actually share their emails with all the patients. Yes. I would love to help with that. Yeah. I think that's a tremendous application. I hear all the time about, you know, there's this term that they use, pajama time, where when the advent of the EHR came about, a lot of physicians ended up just having to do all of this paperwork after their clinic hours, and it really ended up adding to a lot of burnout. Yeah. Which is really unfortunate. I think we need to do a lot more to support the clinicians. It's one of the hardest lines of work. It's one of the most important lines of work, you know, as you feel when you or a loved one is going through a health episode. Yeah. And I think the system makes it really hard on them, honestly. Yeah, like printing out and, like, typing all the stuff. It's, like, very manual process, right? Very manual process. And to the point that, I don't know if you've experienced this, but I've seen in caring for my parents and also even for my own health, a lot of those, they call it, like, in-basket, those emails, you know, in the chart, they're having to have, like, PAs and nurse practitioners help with some of the responses because it's just too much for any clinician. So anyway, big plus one to your idea. I think there's definitely something there where this technology can help them in the same way that, I don't know if it's helping you, but it's helping me with my own inbox. Yeah, like, I don't like filling medical forms online anymore. I mean, obviously medical stuff is more important. You gotta be careful, but, like, I just get AI to do it for me. Yeah. Yeah, I think we're at a point where there's still that need for sort of the human-in-the-loop piece, but it's still a huge time savings to have. I still review all of my emails. I don't have AI agentically sending things out without my review still, but it still saves me a ton of time, even if I end up making little tweaks here and there. So I think that's a good application. Awesome. So another thing that really grinds my gears is, like, maybe I spend too much time on X. But, like, some random would just tweet like, Oh, AI is about to cure cancer, or, like, this kind of stuff, right? Like, AI is, like, very imminently about to cure cancer because some, like, you know, some researchers or something. And I saw Kate, the former CMO, tweet this, saying, like, To actually give frontier cancer care is, like, very difficult from the standard care. Like, have you had—can you talk a little bit more about this? Like, how can people actually even get this frontier, like, state-of-the-art care? Like, you have to spend a shit ton of money and, like, really go out of your way, right? Yeah. I mean, and I appreciate that you and I have the shared experience of actually supporting loved ones through cancer, so you know it intimately in the way that I do. I think there's such a journey, right? It's not just the treat—I know when people say cure cancer, I think a lot of the time what they're talking about is the treatment, like having an actual cure in the form of some sort of, you know, therapeutic or something that can actually fully reverse the disease. But to me, surviving cancer is so much more than that. That's obviously really important. I do think we should point this technology towards science and towards that piece of the problem, for sure, and we are. But there's this whole journey of, like, from the moment that you get the first diagnosis, or actually even upstream of that, the moment that you think to even do the preventative screen, or maybe you experience some symptom that gives you this trigger of, like, maybe I should get this checked out. Like, that's, for me, that was the scariest part of, like, going from that sort of uncertainty of an early sign all the way to diagnosis. You're hanging in limbo with a lot of uncertainty. And then once you get the diagnosis, that, to me, that was an equally difficult part of that limbo period of now I have a diagnosis, but I don't know what the treatment path is, and now I'm evaluating a bunch of that. That was, like, actually the most stressful for me as a control freak because— I felt like I had an ability to potentially change the outcome based on how we steered the treatment path, and I didn't feel like I had full information. So anyway, I say that all to say that I think we should talk about, like, cancer care and sick care and any sort of— Cancer care and sick care and any sort of healthcare as more than just cures. Yeah. So much of it is—I didn't even talk about the preventative all the way upstream, right? If there were things that we could have all done, you know, the first question they ask when you have lymphoma patient is like, Did you smoke? Did you have birds in the house? You know, there's even—so there's a really long journey here. And I think the promise of what this technology—and I think we're already living it in some ways, but there's so much more to come in the future—is really how it can transform that whole journey. Yeah. Right? And so, yes, I do think when you see what's happening with the latest models, what they're doing in the fields of mathematics, for example, you have to believe if we come together as an ecosystem, which we should, to solve these important problems, like some of the diseases that have eluded cures, we can have some new scientific breakthroughs. But I think we would be fools to only focus on just that piece. We should focus on the whole. That's a really good point, because actually the best cure right now is just to get scans regularly and find this stuff early, so you can just cut it out, right? That's probably the one. Yeah, I mean, I'm not a doctor, but I do think that generally prevention, I think, can be an antidote to a lot of things, not just cancer, but all sorts of—you know, actually some of the largest determinants of death, there's like cardiovascular issues, metabolic issues, and a lot of those are actually things that don't even—it's like getting better sleep, like changing your diet, like moving. Everyone knows kind of intuitively that those are things that you're supposed to do, and yet it's still so hard for all of us, myself included, to adhere. And that's actually one of the things I think is really cool about AI. I hear these stories. I heard from a woman, for example, who was postpartum, like a couple years postpartum, and she just really struggled to shed what she called baby weight. And in her opinion, she just had, whatever, 15 pounds that she just really wanted to shed because she just didn't feel like her best self, like her pre-baby self. And she tried everything. And she wrote in to us saying, when ChatGPT on a whim decided to just kind of confide in ChatGPT, I was like really feeling down about myself. And that was the moment that it clicked for her. And it was purely because this tool that she had felt like a no-judgment zone, and she felt like she would be honest. Yeah, that was the difference. And there was some day-to-day tracking. You know, she was sending progress photos, she was logging her meals, all those sorts of things. But to her, you know, there were other products that she had tried doing that and didn't work. For her, the difference was having the space where, like, you know, there was a day where she would have an off day where she didn't feel like doing the workout, and she didn't feel embarrassed to confide in that. And Chat's response was, you know, get back on the horse. It's okay. Maybe it's just a 10-minute walk today if you don't feel up to the, you know, 30-minute HIIT workout. It's okay. Like, think about the long journey. And it's almost like that support. So anyway, I just, I think this is a much broader conversation than just, like, the moment you're in a point of care. Yeah, because, like, you know, ChatGPT has perfect memory, right? Like, if my wife complains to me about something, like, I'll forget about, like, if she was, you know, she was saying something like a couple days ago, but ChatGPT will know. So it's kind of like, yeah, yeah, definitely should use these tools. And that's a good segue to talking about, like, how regular people can use ChatGPT health. I think the first step is just to, like, connect all your accounts, right? Is that the first step? Yes. Yes. If you think there's kind of a vision here where you could imagine the sort of best nutritionist, fitness coach, advocate all in one, that's sort of the promise of what Chat could do for you, all the way from sort of prevention to if you get into an episode care, what you and I do for our parents. This technology could do that for you, even to booking appointments one day. But in order for it to do a good job, in the same way that you and I need to understand the details of our parents' cases, it needs to have that context on you. And so step one is just we're trying to make it really easy for you without having to, you know, do the old-fashioned way of, like, uploading tons and tons of, like, finding all your labs, downloading, uploading, what you're having to do with your mom and the Google Drive. It works, but it's very manual. And so what we've done is made it easy to just single tap and put your login information for your health systems, connect your health records, connect your wearables if you want to, and then have that synchronously going so that if you go to see the doctor or you take some steps or whatever, that's just automatically context that Chat has so it can personalize the answers for you better. Nice. And I feel like talking to my Noom friends, people not in the Bay Area, like the biggest barrier for them to really get the most out of ChatGPT is just, like, connecting this stuff. They're like, Oh, you know, it's going to leak my information, or, like, blah, blah, blah. And I always tell them, like, Hey, this is like a trillion-dollar company. They're not going to leak your information. But, like, do you have any, like, reassurances for people watching this that, you know, they're taking this seriously? Absolutely, yes. It's something that, to me, it's probably the single most important feature that we put all of our energy into building first before we did anything else, is really shoring up that security and privacy approach. Because I too, I mean, I put my health data in here. So I'm eating the dog food that I may, you know, we use that term in the industry, dogfooding. And the reason that I feel comfortable, and I think other people should too, is that we've really built that. layer of privacy and thinking about the security of the data into the way we build the product. There's additional encryption layers on the health data. If you sync your health records, for example, your health data, we're doing extra layers of kind of purpose-built encryption. And importantly, we don't train on your health, the health data that you connect. Yeah, yeah. We don't target ads based on that. So if you connect medical records, we're not training on or targeting on that. Okay. These are kind of design decisions that, you know, we're doing because it's the right thing to do in terms of preserving that trust and building something that people feel comfortable. Yeah, that makes sense. So beyond, like, the standard, like, you know, the steps and you can connect your medical records with ChatGPT, and what if I want to, like, you know, research clinical trials or, like, get into somewhere more like, you know, people with long-term illnesses? Like, is there plans for that or, like, you know, like, I guess how do you best use ChatGPT to find that stuff? Yes, yes. There's so much potential for where this can go. I would say today, where the product is at, the best thing, I think, would be to make sure that Chat understands that that is something that is important to you, which is literally just as simple as telling Chat that I have this condition. This is something where I want to always know the latest. I would even, like, set up an automation. I think we'll get to a point where Chat will infer and just know these things for you. But to date, we still have a lot more that we can do in terms of what the models are capable in terms of proactively telling you things than it won't all just happen right now, even though, you know, we'll get there. So for now, I think one of the best ways would be to maybe set up some sort of automation, like say to Chat, I want you to always be looking for the, I don't know, say you have POTS or something like that, some sort of chronic illness or EDS. You can say, I have this condition. I want you to always alert me when a new study or a new breakthrough or some new finding or some new news comes out about this condition. That it can do very well today. Yeah. There's these websites, right, clinicaltrial.gov or something, like these websites where you can just search for stuff and ChatGPT do it. Yes, yeah. Chat can do pretty extensive web search at this point. Yeah, yeah. I think as caregivers, like just going back to that, I think I actually built this open-source tool called, like, FCancer. I don't want to say the word, but FCancer. And all it does is, like, it just maintains, like, a one-pager on, like, my mom's treatment, and, like, it has all the phone numbers at the top so I can call them whenever I want. Wonderful. Yeah, and it just keeps it updated. And like you said, it keeps it updated, like, in a Google Doc or a site? Right now it's a Google Doc. I could use a site too. You could use a site. But yeah, like, it kind of looks through all the emails and, like, you know, like the notes from the phone calls and stuff, and just keeps everything up to date. That's awesome. Yeah, because otherwise, like, there's so much shit going around, like it's hard to keep track. It's hard to keep track. Yeah. You know, you're reminding me of a story I just heard of someone who's battling a chronic illness, actually, and what she did is she created a site where she has her data populating there, and she's given instructions for, I think she wanted particularly her latest medications to always be at the top, and then her latest lab results next, and then I think after that was, like, latest appointment notes or something like that. Okay. And this is a site that she's given access to just sort of her closest loved ones, so kind of like her care team. Nice, nice. And it's a helpful thing because when you're going from doctor to doctor, a lot of the time they don't have full visibility into another system's records. Sometimes it takes a lot of faxes and it doesn't always happen. Faxes. And so what she's found is actually it's really helpful when she establishes care with a new provider or even if she's just, like, ping-ponging between a couple different systems, to just show them this, because when they're prescribing medications, a lot of the time that's how they catch contraindications. Got it, got it, yeah. Because doctors can't hold all the information in their heads for, like, 20 different patients. Of course. They need help too. It would be insane to expect that they would, and even if they could, a lot of the time, again, they don't even have visibility into the full longitudinal record, cross-system record. Yeah, it's, there's so much that we can do, but I do hear this story a lot from people who are really kind of taking a driver's seat and using Chat in this way, putting all their information in one place. There are some people who actually just will always run any new treatment or any new, you know, protocol that comes from a doctor through Chat just to double check. And I've heard stories of people catching—actually, my former boss, Fiji Simo, has a story that she's told publicly about this, where she actually caught a contraindication right before a treatment was about to be given. And her story was like, the attending was like, Oh my gosh, I'm so glad you said that. Absolutely, we should do this other treatment. Oh, really? Wow. And it wasn't because the attending wasn't smart and wasn't paying attention. It was because, like, they literally didn't have visibility into that part of the record. Oh, so she found, like, a separate treatment that— Yes, the treatment that she was about to be given would have had a really bad contraindication, and she narrowly avoided it because she ran it through Chat and Chat caught it. Somewhere deep in her record, it had that history of— That's crazy. Yeah, the indicator. And I hear a story like this, I mean, that's, like, not a coincidence that, you know, my former boss had that story. Like, I hear stories like this all the time where, like, either someone, yeah, narrowly avoided a contraindication, or they found a diagnosis that had eluded them for years. I hear that version all the time. There's this mom, her name's Hillary Eaton. She just got written up in the Boston Globe, so she's kind of publicly telling the story of her daughter, where years she spent trying to figure out what was going on. She obviously saw these symptoms, but didn't have an explanation until using the power of Chat to figure it out, and now they actually have a treatment path. Yeah, I think similar, like, you know, my mom, like, she had chemo before, so you can only have a certain amount of radiation and chemo. So, like, I have records of, like, you know, whatever radiation she got before. Yes. So, like, in that 15-minute conversation I have with the doctor, hopefully it will come soon, I can actually tell them the stuff in case. Yes. Well, that's great. I hope that AI is helping you keep track of that in your sheet. Those are the kinds of things that I think it's really helpful. They'll ask you in the moment, like, When was the last time you blah, blah, blah? or How much radiation, blah, blah? And you're like, I don't know. You can now ask AI real time. That's been really helpful for me for more basic things than that even. You know, just like, What was your latest blood pressure read? Like, a recent— PCP appointment and my blood pressure was kind of low. She's like, Did your blood pressure tend to run low? I was like, I don't know. Let's ask Chat. Here's the last three readings. I just think it's so—it's just so helpful to have an encyclopedia on yourself. And let's talk about, like, ChatGPT for clinicians and physicians. Yes. So maybe we can build some empathy here, because part of me is like, why are these guys still using faxes and, like, these 20-year-old systems inside hospitals? But maybe there's some regulation or something going on, or, like, you know? Yeah, there's a lot of regulation in healthcare. Yeah, yeah. A lot of empathy for these folks. And it really comes down—you know, they're trying to uphold privacy commitments that they have to you and me. And so it comes from a place of deep values and principles that I think any of us would be very aligned to. Okay. Unfortunately, the reality is that what ends up happening is that it makes the whole system a lot more difficult to navigate. Like, for example, when my dad was first diagnosed with lymphoma, you get CTs of the lungs. I wanted to get a second opinion. I won't name the names, but there are two institutions—it's really not specific to them—two top institutions in the Bay Area. And in order for me to get the second opinion from the second institution, I literally had to drive to the first one to physically pick up CDs. Yeah, yeah, exactly. That's what I had to do too. You have to do this too, yes, which is crazy, and then physically bring them to the other institution. I don't even have a device that uses CD anymore. Like, how are you gonna— It's wild. And you know that's not a technical limitation, right? Like, there's clearly a privacy thing. So I think... I think this, I hope that this can be a golden age. And what I'm seeing from every health system that we work with, and we're working with some of the top institutions in the nation right now, there is a hunger for—they don't like this either. You know, it's not—they want to be able to uphold the privacy commitments without having to deal with antiquated technology. Like, they're people too of the world. They, as consumers, get access to all this technology. They're super smart, and they really deserve better. So I do think that we're on the precipice of a moment where AI is—and you're seeing it with some of the ambient scribe adoption. You know, we partner openly with great providers in the space like Abridge and Ambience, who power a bunch of the ambient scribing. You're already seeing, like, this rapid adoption that previously did not happen. The latest AMA study, I think, showed something like over 70% of clinicians are using AI regularly now, and that was up from, I don't know, it was something in the 30s or 30% or so back in 2023. So it's growing quickly. It's happening, and I think faster than what everyone tells me. I've only been in healthcare now for a little over a year, but for people who've been in it for decades, they're telling me that this is unlike a pace that they've seen before because there's just so much to fix, and this technology is actually able to make a difference right now. And as long as, like, if I'm a doctor, I want to adopt this stuff, I can just start using it? Like, it's just like trying to get enterprise adoption of AI. You have to go through a lot of different other parties, right? So there's a couple different paths, which is exciting. So we have a product called ChatGPT for clinicians, which is free for any clinician. They just have to verify that they're a clinician, so they use their NPI and ID, and then they get free access to our latest models. Like, they get Astra for free. That's something that we just feel from a mission perspective is very important. It's a no-brainer for us to comp that for the physicians, so they can do that. Now, for many of them, they might not be authorized to sign a BAA on behalf of their organization. That's a bunch of industry lingo, which basically means that in order to be HIPAA compliant, or in order for us to be able to support their HIPAA compliance, they have to sign this document with us with sort of a legal agreement. So those individual—a lot of those individuals... Individuals won't be able to, for example, sync up all their patient EHR records, is kind of the implication of that. And so it's still a very powerful tool. They can ask questions in a very anonymized way. You know, I have a 73-year-old patient that presents with X, Y, Z symptoms. I'm thinking about blah, blah, blah as the causes. Like, give me a sense of what your differential would be, and it just helps them as almost a thought partner. But it gets even more powerful when it is able to integrate with the electronic medical record information and be in workflow. And so that's why we're also working with health systems directly to then actually integrate, you know, in a HIPAA-compliant way with their systems. And that is pretty consolidated, or is it pretty fragmented? Is there like a couple electronic medical record platforms, or— There's a handful. And the thing that's nice about, you know, the health systems sort of write to their own data is that we can work directly with the health systems and work with them and whatever electronic record they're working with, we just work with them to say, what's the way that you would like this to work for you? And we figure out, you know, there's kind of open standards and things that we can use to make those things a reality for them together. Got it. Okay. And how about, like, cutting-edge cancer research and stuff like that? Like, is OpenAI—I'm assuming you guys are helping that too. We have a large emphasis and focus on science and advanced— and focus on science and advancing science with our models. It's some of the most—it's very exciting work, both through actually our PBC and our foundation. Our foundation has publicly shared that a big portion—I'm going to botch the stats, I don't know exactly, but we can look it up after—a big portion of those funds, a lot of money that the foundation is deploying, is dedicated to advancing life sciences. Okay, okay. So there's kind of two angles. Yeah. And it's a huge priority right now to say, how do we work together with the ecosystem to accelerate discovery? Okay. Yeah, I'm really excited for that. Me too. I think we're talking before the interview, but, like, you know, everyone's building personal agents and stuff to help you manage your emails and, like, this kind of stuff. That's great. That saves me time. But, like, to actually improve human life, like, you know, cure these freaking diseases, like, that is massive improvement. It's incredible. I mean, I think all of it's incredible, to be honest. I mean, we have 300 million people coming to Chat every week asking health questions. Like, Chat is supporting 300 million people in their health already today, even without the acceleration of the cures for the whatever. Yeah. Which I think is just amazing. It's already happening in front of us. And we're also seeing these models that have this next-level intelligence that are solving, you know, decades of math, right? So it's like, just imagine what's going to happen when we come together as an ecosystem with all the parties involved to point that towards cures. And importantly, there's also, you know, a lot beyond the discovery of the cure. There's also a bunch of steps after that to actually make sure it gets in the hands of people, you know, everything from clinical trials and deployment, right? And what we've seen, you know, we work in partnership with a lot of players in the space, pharma companies, et cetera, and what we've seen is that there's a tremendous— Potential for AI to accelerate every step of those processes, yeah. Which is also really exciting. So it's, again, I just, I—the thing that's so cool about this technology, yeah, and why I'm able to have the impact that I'm able to have even a year in here is because you can apply it to basically every part of the problem, yeah, and it's able to benefit in some way. So it's pretty exciting. Yeah, there's like a—like, because we both work at Meta, right? And, like, part of me is like—part of me feels like working in healthcare, there's long shipping cycles, and, like, because, like, at Meta we can just run an A/B test, change some algo toggles, and you can see the result in, like, the next day. But do you feel that now you have to wait longer, or do you feel like there's, like, a lot of progress making? Well, it's yes and no. I mean, I think we're having some of the fastest progress that we've ever seen in healthcare specifically. Yeah. Now, healthcare has not historically moved as quickly as the consumer tech industry. Social media, exactly. I think for good reason, because the stakes are a lot higher. That's right. And so I think the pace is extraordinarily fast. Now, relative to consumer— Product, you know, social media or whatever, it's definitely slower. But I think rightfully so. All the things we're talking about, getting security right, getting the privacy approach right, getting the UI in such a way that, you know, my mom can figure out how to connect her medical records without having any technical experience. Like, these things to me are so important. It's a little bit of a different approach that you want to take to this type of work, I think, than maybe historical or just traditional direct-to-consumer product. Yeah, you don't want to run A/B tests on, like, treatment and stuff that, you know, or, like, you know, it doesn't make any sense. Yeah. Awesome. Well, I think for people listening to this who are, like, caregivers or patients going through some of the stuff, like, do you have any kind of, like, what's the step-by-step advice that you have for them? Oh, good question. Okay, so, as a caregiver, this is actually an area that I think there's a tremendous amount of opportunity for more, for the technology to do. But at least as this product stands today, my tip would be, or at least what I do—I like your setup, by the way. I think your setup is a great example. I don't know if you do this for your shared project, but I've made the project that I use to manage my dad's health a siloed project, so it's like a project-only memory setting. So when you set up the project, you have to do it, I think, when you set up the project. But if you go into sort of the top-right drop-down of settings, you toggle them out, because most projects just generally just memory is shared across your whole account. Why I do that is it's just kind of like a nice hard barrier to make sure if I'm talking about, you know, the latest diagnosis or the latest lab test or whatever, there's no chance then that chat will get confused that that might be mine. Oh, I see. Because I also talk about my own health with Chat a lot. Got it. So I think that's, like, that's helpful. I hear this also from a lot of parents that are trying to get advice on raising their kids or whatever, that this is a helpful sort of power user way to make sure that that information is kind of... Kind of separate. Okay. Yep. Okay. Just a stylistic thing. I don't know. I prefer it just to make sure there's no chance of the mingling of memory. Yeah, yeah, that makes sense. And then I think what you're doing is a great idea because, yeah, because of privacy and all that, that's probably the fastest way you could have a nice knowledge store that's specific to a loved one, obviously with their consent. Get their consent. I get my dad's consent, and sometimes, you know, if he doesn't feel comfortable, I'll kind of redact, personalize things if that's what he wants for my account, whatever. Okay. Put the latest lab report in or whatever, and I use it all the time. The most actually, and, you know, to help me understand what's going on. Yeah. The most useful thing, though, I've found is even just real time when I'm in appointments. Yeah. Like, as the doctor's sharing things with me, a lot of time I'm using it to sort of real time, like, what does this mean for Dad? Like, what question should I be asking right now? Oh, like in person, you just type something? Yeah. Well, a lot of time, there's a lot of telehealth these days. Okay, okay. So, but yes, I would do it in person too, but a lot of the time we're dialed in. Got it. And I'll be kind of, you know, they'll mention a thing, you know, maybe a latest treatment, or they'll mention some word that I don't know, some acronym. Exactly. Like, you learn so much. There's so many terms of art that you wouldn't know. Like, what's a left wedge resection? You know, I didn't know when we were talking about lung—that's a type of surgery—but, you know, it was really helpful for me real time to be asking, and so I wouldn't have to waste the time with the doctor. My dad would be asking questions, and I'd sort of side be educating myself, and then I could make better use of the time. Got it. I don't know, what tips do you have? You've been doing this a lot. I think if you actually, like, a lot of people when they have these problems, they don't like to share that they have cancer. But if you actually share, or maybe because, like, maybe I share too much, a lot of people actually come out to help you. Like, I'm in touch with an oncologist in New York who's just giving me free advice, and a lot of people just come out of the woodwork to help me, which is very nice. That's amazing. Yeah, people actually do really care. That's a really good point. Like, lean on your community. That actually, that is a great direction to take this because— I am the biggest supporter of what this technology can do for us, and yet still, there's nothing that can replace human connection. And I think it's important for us to, as we have these conversations, talk about, in our best case scenario, what these tools actually do is they strengthen the human connections that we have because they allow us to show up in such a way that we're able to, you know, help each other, help ourselves, whatever, such that we have more of that time for human connection. I totally agree with that. I think having a community when you're going through—like a lot of your friends might be going through the same thing as me. I'll be talking about it. Like a friend called me, he's like, Hey, you know, my dad went through the same thing. Here's how we did it. And it's like very—you know, you don't talk about this in regular life, right? So, like, it's good to just share. No, it's so true. And you don't know until you share. I had a similar situation with my dad—the second cancer that my dad had. One of my best friend's father-in-laws just happened to be in the space and was able to—they live in Pittsburgh, so they're kind of far away—but the doctor was able to refer me to a specialist close by us that worked at Stanford, who ended up being my dad's doctor for that other cancer. Got it. So you're right, you just never know. And you're right, people really do come out of the woodworks to help. Yeah. Yeah. All right, let me ask you one last question, which is, I've always found it, like, as I got older, I've always found it difficult to balance my career with actually spending time with my parents because, you know, your parents aren't around forever. And, you know, I think working at OpenAI is a pretty demanding job, right? So how do you think about this from your own point of view? Oh, man. Yeah, this is a tough question. Yes, this is the age-old question. It's hard to do it all. I think I really do think—so there's all the, like, optimization hacks and trying to get the most out of every minute that you have and blah, blah, blah. But at the end of the day, we all have only 24 hours per day. That's right. And so there's only so much you can max out. I think it really comes down to being really clear about what are your priorities and if something's going to drop, what, realistically, what's the stack rank of things? And I think it's going to be different for everyone, but it seems like you share a value system that I have, which is that my family is the most important. And so for me, it's just a non-negotiable. Like, if my dad has a doctor's appointment— Yeah. Even at this job, even in my former job at Meta, I'm lucky I've worked at employers that are understanding and supportive of these things, but, like, I also still have to set that boundary. I'll just block the time. Yeah. I just block the time. And not once have I ever gotten— Pushback or criticism or questioning. Everyone's so supportive, but there's something that has to flip in you because until my dad got very sick, you know, this was back in university, I still, like, ran kind of this crazy sort of always-on life, even as a student. I always felt like there's no more time in the day. There's no more possible thing I could do right now. And then someone gets sick and you realize, wait, actually, when I really prioritize this, like, look at what I'm carving out and, like, what could drop. And, you know, it's not always great. Obviously you'd like to do everything, but I think sometimes when you prioritize, you realize, like, how much you really can dedicate time to certain things if you just make the hard calls. And are you just limiting yourself? Like, are you deluding yourself that you're actually that busy? Maybe some of those meetings didn't actually have to happen, or, you know, and speaking for myself, you know, like a lot of PM meetings don't have to happen, so I don't think. Yeah, yeah, yeah. Or something doesn't need to happen right then. Maybe something can, like, you know, not everything. Some things are time-sensitive, but sometimes there's, like, self-imposed, you know, and some things actually benefit from an extra week of breathing for people to come together. So I don't know. I think I had a similar feeling when I had kids. I have a one-year-old and three-year-old right now, where you just start to realize that you can... You can make time for the things that are important if you really prioritize them. It's really that simple. You have to, like, kind of, like, not take your loved ones for granted, right? Because you assume they'll just be around forever, but your kids will grow up and, like— No, you have to prioritize. Yeah, you have to prioritize. I know. Do you have any tips on it? Well, I think, like, I kind of made the conscious choice to, like, for finality, to kind of, like, focus on my mom's care and, like, make my little YouTube videos and stuff. But, you know, that's kind of, yeah. Great YouTube videos. No, I totally agree. I mean, you're actually reminding me, there was a period during—it was around the pandemic—where both of my parents each got diagnosed with two cancers, each within a week of each other. Oh, shit. And I had just had a miscarriage, and it was just a tremendous— Wow. —like, a really hard month. I actually took that—I, we have this program, or my former employer, we had a program called Recharge, which was sort of like every five years you get to take a month, and I used—and I just took the month. I did a similar thing where I said, okay, this is, I'm gonna dedicate this time. There's no shame in taking slowing down your career. No. It was probably the best, one of the best—I don't regret that a single bit—but it was one of the best things I could have done in that moment to just both care for my parents, but also you have to care for—I guess that's the other piece of advice—you have to care for yourself as the caregiver. Yeah. We don't talk about that enough. It's a big, even with the help of AI, it's still a really big cognitive load, emotional load, just pure time load to be a caregiver. And you have to also make sure that you're budgeting time for your own— Your own mental health. Your own mental health. Yeah. Yeah. Yeah, I think AI and talking to your community helps with that. I think so too. And I think it can make—you just feel just more empowered and being able to— Yeah, like not alone. —take on that job. Yes. I feel way more empowered now to do that job for my family and for myself than I did without Chat. It's amazing. I'm just so excited for where this can go because you're mentioning several times, you've mentioned some of the administrative stuff that it seems like things are coming back to. It's—I agree, and— Just with where the technology is going, it's just, it's really exciting to imagine a world where, you know, Chat just knows that, like, I just got a PCP appointment. I'm supposed to do follow-up labs. Chat could, the technology is totally capable now of just letting me know, Hey, I saw your after-visit note. You've got these labs. They're supposed to be through LabCorp. Here's an appointment and a time that works for you. You're working from home this day. Here's the time slot. Do you want me to book it for you? And I could literally just push a button and it could make it all happen. I mean, that's the future we all deserve to live in. I'm so glad, actually, that you're working on this with the team, and I really think this could be the biggest positive impact on humanity. So thank you for working on this. I really hope so. Thank you. Please send me your feedback, and I'm just really excited to keep making it better for folks like you.