Transcript
0:04 AI has created an existential moment for humanity. Right now, the conversation is overwhelmingly driven by questions around AI safety guardrails needed to minimize the risks. But we need to use this window to expand the conversation. It can't just be about how to guard against the risks of the machines. It has to include how to make humans more capable. AI is surpassing humans in intelligence. For centuries, we've defined ourselves by our intelligence. But if, "I think, therefore I am," no longer defines us, what does? If it's not intelligence, what is the one thing that makes humans uniquely human? By dethroning human intelligence, AI is forcing us to rediscover our consciousness, our soul, our spirit at the center of our humanity. This moment, if we take advantage of it, far from exposing our limitations, is revealing our limitless possibilities. But while we are investing staggering amounts of money, time, and attention on making the machines more powerful, we're devoting almost no resources or attention into making humans better. Why do we see machines as dynamic and endlessly improving, but humans as static and stagnant? As we marvel at the progress toward recursive self-improvement in the machines, we've lost sight of the recursive self-improvement that's built into our humanity. Our fundamental drive to grow, learn, improve, and evolve. What AI becomes will depend much more than we realize on who we become.
1:52 In short, AI needs a new story, a bigger narrative that touches people's hearts and minds. So far, those building AI have been speaking mostly to a small circle in technology, business, and media about singularity and exponential. In this next AI chapter, they need to speak more to the second audience, which happens to be the rest of the world. If we're going to make the most of AI, humans have to improve along with the technology. The biggest existential risk is that we will place all our bets on the machines. To help tell this new story, there is no one better than Daniela Amodei. Daniela, thank you so much for being with us at Think Health. Think Health is, of course, founded by Alice Walton, who famously brought the Crystal Bridges Museum of American Art to the Heartland, who built a medical school and a Whole Health Institute, and is passionate about the use of AI to improve health. There is no one in the AI world who epitomizes this combination of AI, the humanities, and passion for health and human flourishing better than you. You are at the forefront of frontier models, but also you studied literature. You play the flute. You're a mom. You don't just speak tech. You speak about nature and friendship and self-reflection. So we really look forward to exploring these deeper questions together. Thank you. Thank you so much for having me. It's such an honor to be here. Let's start with this moment we're in.
3:38 Right now the AI conversation has focused on all the risks and the backlash against AI, with 75% of people being fearful or anxious, and not so much on the benefits. And of course, the anxiety is real – about jobs, about safety. But do you think that this anxiety is a proxy for something larger, the fact that suddenly we are not going to be the most intelligent thing on the planet? Is this anxiety about meaning and not just economics? What do you think about that larger existential moment?
4:25 This question is so important because my view of what is happening in this moment is that many people are starting to experience both the power of artificial intelligence and are also collectively really starting to think about what are some of the risks? I think Anthropic is a little bit unusual because really from our inception, we have had this desire to be very open about both. One of our core cultural values is to hold light and shade. And from the very early days we've always said we think this technology is going to be capable of doing incredibly impactful things for humans everywhere on the planet.
5:15 But there are real risks that we need to mitigate in order to be able to get to all of the incredible and positive benefits that we think AI will bring about. Anthropic has actually pioneered many of the safety techniques and talked publicly about the need for common sense regulation. This is everything from our work in interpretability and constitutional AI to our work in safeguards. This has always been important to us to show that we take the risks seriously. At the same time, part of why we chose to embark on this journey of making Anthropic is not just to talk about the risks, but because we believe that the potential benefits, not just in the future, but today are very, very, very large from artificial intelligence.
6:07 And we've talked a lot about what we hope to see in the future. Dario, who's my brother, our CEO, wrote this very long essay called “Machines of Loving Grace” where he talked about a future world of massive improvements in biology and medicine. While that is an incredible world that we are building towards, I also sometimes think we could talk a little bit more about the benefits that people are really experiencing from artificial intelligence today.
6:41 Part of what is so unique about this moment is that a lot of these benefits are happening and are a little bit quiet. I think a great example is from one of our partner organizations called OpenEvidence, which uses Claude to help doctors access all of the medical literature in the world. So whether you're a doctor operating in Cleveland or in Ghana, whether you're in a city or in a rural village, you're able to access the entire scope of human knowledge and innovation about whatever medical challenge your patient is facing. You have been very unusual among AI leaders in speaking a lot about wanting to be self-reflective, about learning from mistakes and about nature, about friendship. And that's kind of unusual, because in our social media ecosystem, we are kind of prevented from that self-reflection. It's so easy to spend our lives doom scrolling or losing ourselves in the rabbit hole of social media.
8:53 So how can we help people nurture that part that wants to be self-reflective, that wants to reconnect with themselves and what really matters? And can AI, with its power of hyper-personalization, help with that? This is such an important and a very astute question. My personal view is that we're coming off the back of this time in history where technology in the social media age has largely been used to disintermediate humans from one another. Instead of you and I going to have an iced coffee together in person, we're messaging on Snapchat or Instagram, instead of people going and forming communities in person with the people that live in their neighborhood, they're writing angry comments.
9:54 Anonymously, often. Yes, anonymous comments that an algorithm served them to manipulate them to feel mad so they can continue to be on this platform. It's for another day to kind of diagnose all of the reasons that led to this particular moment that we're in. What it's led to is a warranted skepticism from people that technology can be used to bring us closer together, because I believe that was the promise made by these technology companies.
10:25 And broken. That essentially broke those promises. Exactly. So I understand where the skepticism comes from. And I think in a lot of ways it's very warranted. At the same time, the way that I think about AI is quite different. AI is not a social mechanism. There are some applications of AI that we think are quite bad, like making AI girlfriends or boyfriends. I think that's a terrible use of AI and people should not do it. And the reason that I think that is because it is another step further down the road of disintermediation, of humans being further from one another. A much better world of AI is one in which artificial intelligence is doing a lot of the things that are isolating or lonely activities for humans. They are making us more productive. They're doing work that an AI can do that a human doesn't have to do, so that we can spend more time together. So that a doctor, instead of spending time – ten hours, 15 hours – researching what is happening with this patient, they can use OpenEvidence, to go back to that example, and say, I actually have access at my fingertips via Claude and OpenEvidence to ask what is happening to this patient so that I can actually spend more time relating to them as a human? I'm not doing as much of the diagnostics myself as a doctor. Potentially, in the future, I'm spending more time actually relating to and empathizing with my patient, understanding all the systems that are going on in their life that might have led them to be sick, helping them feel better and talking them through what their treatment plan is.
12:03 That's just one example in the medical world. But you can also imagine that socially, it would probably be a better world if we were spending less time interacting with technology and more time interacting with each other. And also connecting with ourselves. I studied comparative religion in India and what amazed me is how every spiritual tradition, every religious tradition in the end says the same thing: that we all have this place of wisdom, peace and strength in us. But our current world makes it very hard to reconnect with it. Thich Nhat Hanh said it's never been easier to run away from ourselves, so how can AI be used to help us reclaim time, as you said, but not by using that time we found in the same way we used to use it? As you know, we have a lot of data now that a lot of people are saving time by using AI that makes them more productive and but then are using that time to do the same things they were doing before.
13:09 So there has to be a process of educating ourselves, nurturing that part that we all have in us to reconnect with something deeper, both within ourselves and with each other. I think this is so, so accurate. Again, not to overly pick on the social media companies, but I think in addition to disintermediating us from one another, it has also in some ways disintermediated us from ourselves. Yes.
13:40 Instead of spending that time with other people, that's also time that could be spent in nature with animals, in self-reflection, taking on passion projects. Part of what has been really inspiring to see about some of the use cases of Claude and artificial intelligence is the possibility for it to spark creativity, curiosity and joy in people. The ability for people to really be able to access these tools and learn – we were talking before about how we both love Greek mythology – from talking to Claude. It's a tool for me to also explore what it is that brings me value and meaning. I ask Claude to help me pursue many of my passions.
14:26 Either by giving me more information by recommending things, by helping me create projects that are valuable to me, even if they're not for productive work reasons. An example I use – I'm not a developer, which I know is unusual for a technology leader – but I've always wanted to be able to build a website. I have a really creative side that usually looks much more like music or art or literature, but there's something really powerful about expanding the ways in which you can be creative. And my ability to just do things for fun and enjoy myself has increased with being able to understand and have fluency in these tools. I don't think those are the only applications of them, to be clear, but I think there are universes where that can be one of the ways they provide value to people.
15:17 You've said that in a world where AI is every day becoming more and more capable, the uniquely human qualities are going to become more and more important. So what do you think are some of these uniquely human qualities and how can we nurture them in each other and in ourselves? I’ll use Anthropic as a little bit of a test case. When I think about employees at Anthropic, just like people everywhere in the world, we have some nervousness. What does it mean if I'm a developer and Claude Code is so good at writing code? What does that mean for me as an engineer?
16:02 A lot of what we found is that while these models are getting very good at doing these tasks, even end to end, what that has meant for our engineers is that they have a lot more time – yes, to continue to do architecting and coding and working with the agents – but also to spend with one another and with other people in the company and with customers. So engineers might go from maybe 40% of their time hands on keyboard to 20%. What they're doing is redirecting that time towards building better relationships with product managers, listening to customer feedback, thinking about how to bring more joy and delight to the products we're building for people.
16:46 It's interesting to me that when we have more time, we often tend to redirect our energies towards other people, or towards learning ourselves, right? Some of these engineers have said, I'm actually really curious to know what it means to be a recruiter, or what it means to work in sales, or what it means to be a product manager. I think the ability for people to learn more has really gone up. My belief is humans are genuinely, genuinely curious. We have a sort of instinctive desire to want to learn, to be creative, to be curious.
17:25 And to grow. To grow, to develop, to help be the best versions of ourselves and help other people be the best version of themselves in kind of the purest form. It's obviously not a panacea, right? It's very complicated — AI has a lot of risks but a lot of benefits. If we're able to get through the challenges, my sense is that on the other side is this world in which human skills are the main thing that's left.
17:56 If you imagine, I'll go back to my doctor example, you might today go to a doctor who you want to be the best, most educated, best diagnoser of whatever. You come in and say, oh, my head hurts. Why does my head hurt? And they're like, here's all the medical things that could be going on. But in the future we might optimize for doctors who have a great bedside manner, who care about you personally, who show they care. There's so much interesting research showing that clinical outcomes are better when you have a good personal relationship with your doctor.
18:29 That's just one example. Imagine teachers not spending as much time developing curriculum and on all of the logistics and administrative duties that come with being in a school. Maybe they're actually able to spend more time with kids and spend time doing what they love: relating to young people, helping teach them important values and lessons, connecting to why the discipline that they chose is important and meaningful to them. I met a woman recently who told me she runs a moms group for moms who use AI. She said that we carry the majority of the mental load in a family and the ability to take some of that mental load and give it to Claude enables us to spend more time with our families, with ourselves, with our friends, pick up passion projects, and be curious about things again. This is a world that we all get to choose to build together. This is something that everyone has a right to participate in and say, how do we want to not feel like this technology is happening to us, but that we're able to take it by the reins and say, how can we use this to make it work better for ourselves, for other people, for our society? Because I think a terrible outcome would be one in which we are further disintermediated. A much, much happier, better world in which we are spending more time with ourselves and with each other is a world that I think we all want to build right now.
19:59 I love how your whole life before Anthropic brought you to these realizations. Because before Stripe, before OpenAI, before any of this, you were a classical flutist on a scholarship. I love that because music demands something so different from us. You know, just practicing for hours for a performance that's not going to be rushed and that's in the service of something beautiful rather than something useful. What did you learn from playing the flute that has helped you in being the president of Anthropic? Interestingly, a lot of the themes that we've been talking about are quite resonant with this. I personally loved the structure and discipline of practicing. I've always been someone who likes to work hard, but I think the ability to work hard at something and have it be a creative and meaningful and sort of cup-filling outcome — it's a very special privilege to be able to work hard at something and feel fulfilled from it. I think I first got a sense of what that felt like from music.
21:14 The second is that the part I enjoyed the most was playing with other people. And so in college and even before college, I did a lot of chamber music and I played in an orchestra. Part of what was so exciting about that, it sounds almost a little bit trite, but you're so much greater than the sum of your parts, right? You have every different instrument that has its own special role. Everybody has the moment where they're in the spotlight and the moment when they're in the background, the moment when they're supporting another instrument, and the moment when they're taking center stage. And I actually think about that a lot in how we build teams at Anthropic and how we work together.
21:53 There's a moment when it's one person on my team's job to be in the spotlight and somebody else's job is to support them, but that changes and comes back around all the time. And it's true that each different position, each different instrument – the shape of what the spotlight means to them – might change, but it is absolutely the case that we're stronger together as a team and that things feel more meaningful when we work together.
22:21 And I think it's not talked about as much, but it's such a privilege to work with people who are so talented, so passionate, and who care so much about the mission, and get to be the conductor and decide how we want to structure this orchestra so that everybody at Anthropic can be their best selves, be curious and learn. The last thing I'll say is there is a particular experience, especially when you're on stage with many other people, where you can viscerally feel this emotion that you're part of something bigger than just you. You're part of this whole unit. You're producing something. I still love to go to the symphony and something that's really valuable about those moments is that they're human, they’re creative and they’re ephemeral. Even if you record them, it's not the same experience as sitting. I love listening to music online and streaming, but there's something about the experience of sitting in a theater and hearing people in this moment that will never happen again that I think touches something very deep and human in you. I am grateful that I have an appreciation of what that experience is like from having played music from a very young age.
23:41 I just love how extraordinary life is. Who could have predicted that somebody goes from being a flutist in an orchestra to having everybody report to her in the biggest AI company in the world? Could anybody have predicted that? You already mentioned my favorite Anthropic value, which is holding light and shade. This is so key because it's not just true about AI, it's true about all of us. We are all a mixture of better angels and lesser angels. So we are now at this historic moment when we talk about how AI can be constantly improving, but what can we do to also nurture our better angels?
24:32 What can we do as a culture and what can AI do? This is actually advice I give to the company a lot. Anthropic is a very quickly growing company. We feel the stakes of the mission very much here, and something I tell people is my belief in the importance of balance in life. I love my job, I work really hard and I'm also a mom. I have a five-year-old and a one-year-old and they are on their own schedule that is totally unrelated to my job at Anthropic. To me, the best version of life is one in which humans get to experience a lot of different facets of life that help them, like you've said, self-reflect on who they are, build community with other people, be curious and be creative in ways that are meaningful and spark joy for them.
25:24 And I feel very fortunate to be in a position where I can do a lot of those things a lot of the time. And something I tell the organization is time off and rest is really important. Yes. I find it's not just for enjoying yourself, although that's really important, but for the ability to give yourself space, to be curious, to be creative, to just follow your passions – whatever it is that you're interested in. So many decades of psychology research and anthropological research indicate that there's a few core things that help humans feel like the best versions of ourselves. Right? One is adequate time to spend with the people you love – your family, your children, your friends, your partner, anybody that's meaningful to you. Like you've said, time with yourself, reflection time – for many people that’s religions, spirituality, time in nature, time feeling like you're connecting to something that's bigger than you. Work and meaning is another major area that often is self-reflective for people. And the ability to be able to spend time in nature or walking, exercise — this is not anything that's new. These are reasonably well documented as things that help people feel like the best versions of themselves in some form.
26:45 What's interesting, we were talking about this, with social media is it really cuts against almost each one of those things. It makes it harder for us, it sort of gives us the cotton candy version of connection. It tastes really sweet for a minute but it's not actually filling. It makes it harder to be moving around and sort of be embodied. It makes it more challenging to do things that are meaningful, creative, or make you curious. My hope again is that AI – there's a possibility that we could reverse some of that trend. The idea that AI could give us time back at the sort of tasks that we do not feel are valuable for us to do, whether that's balancing the mental load if you're a mom, or doing technical research and diagnostics and paperwork if you're a doctor. If you're not doing that work, what does that open up time for you to be able to do? Spend time with the people you love in person. Spend time in nature. Spend time connecting to something bigger than yourself. Being curious, being creative. I think this is a decision — we all get to decide if this is the type of world that we want to build. But also, as you have said, companies have to decide. You made some very key decisions, like not allowing ads because ads require engagement.
28:02 And the easiest way to engage people is by activating their rage and their anger, their lesser angels. So these are also very important company decisions. But you've also said when asked about what's the good life, as Greek philosophers always asked each other, and you said it's really how do we become the best version of ourselves. But at the moment we seem to be placing all our bets on the machines. We are placing all our bets on the machines getting better and better and better. It's as though we think, human beings are static and stagnant and machines are these amazing creatures that have recursive self-improvement and we are awe-inspired every day by their feats.
28:58 But we all know from science and from spiritual traditions that human beings also have recursive self-improvement. You've alluded to it many times when you said we're curious, we want to learn, we want to grow. I really believe that this instinct to evolve is a real instinct. I wrote a book many years ago, don't worry only three people have read it, called The Fourth Instinct about our instinct beyond survival, sex, status, power – our instinct to evolve, to find meaning in life. And I wonder, how can we open up the conversation to include recursive self-improvement for humans? I absolutely love that question. To me there's almost three phases of it. The first phase is almost like a base-of-the-pyramid principle, which is we can't be in a position where we're turning that impulse off, which is my worry about the current mode of communication of technology through social media.
30:12 And you alluded to this no-ads commitment from Anthropic. That is to me part of the base layer of the foundation, this idea that if you structure AI products the same way that you structured the last generation of products, you will wind up with the same results. And I actually think it's worse because AI is so personal. And so much more powerful. So much more powerful, so much more personal. Ads targeted at sensitive health data or at people who have emotional conversations with AI – they talk about their family, about their kids, about tough things they're going through at work or with faith or whatever it might be for them – feel disrespectful, advertising to people based on the kind of content they're putting into these tools. Maybe even more importantly, just to your point, the incentives that it sets up is to keep you on the platform, to keep you engaged, to keep you frustrated, and to keep you on a screen away from humans, away from connecting.
31:22 I think if we do things that way everyone is really going to regret it, which is why we made this pledge and which drives a lot of our thinking about how we actually design our projects. Does Claude end a conversation with you when you've gotten the information you need? Is Claude's goal really to be sort of friendly but professional? Not to build a relationship with you because there's no incentive to stay on the platform for that long. I think that base layer we absolutely have to do. The second is really about if the technology is doing its job the way it should be, how do we reclaim time and space to have the ability for people to recursively self-improve?
32:00 Right now, I think so much of the challenge is that modern life is very busy and it's very stressful. And so I think, again, this world in which artificial intelligence is giving us the ability to do a lot of things that humans don't necessarily have to do, if we can use some of that, if some of that need can be met by artificial intelligence, humans will have more time for all of the things we've talked about – for each other, for themselves. That, to me, is the second layer that gives people agency and time to be curious, to self-reflect, to be better versions of themselves.
32:38 And then the third, the very top of the pyramid, is how can we actually leverage these tools to drive people's curiosity? This is a little bit more speculative, but the idea of like, you're really interested in creating this art project. Let me help work with you on how you and your friend are going to do that. How do you actually get the permitting to be able to do this in a public space? Okay, let me help you apply, let's do the application so that you can actually build this mosaic in your town square that you've been wanting to do for 20 years with your best friend. And so again all the first two layers have to be there. But I think there's a role for AI at the top as well. It's so incredible to have an AI leader like you talk about humans recursively self-improving. Can you please be talking about it every day?
33:25 It's so important because it really validates what you said earlier, that bringing AI and humanity together is going to be what we need for a great future. That we're better together than growing separately. And it's a little bit like a dance, or a marriage — you're married, I'm divorced — but we all know that if one partner keeps growing and getting better and the other partner, in this case humans, stagnates then it's not going to be a successful partnership. So how can we – through expanding our conversation, through investing in different institutions, whether it's preschools, colleges – how can we keep helping humans become more and more capable so that this dance in this revolutionary era can be an amazing dance rather than a destructive dance?
34:31 It's so important to have this conversation. How do we help humans come along in the era of artificial intelligence? My framework for thinking about this is the whole reason we're developing this technology – to help humans flourish and thrive and live better lives. So in order for us to achieve our mission, we have to think about how do we actually do that practically? How do we make the tools such that they promote people being able to do all the things we talked about – to have more time, to be more curious, to live healthier, better lives wherever they live in the world, to increase access? This sort of comes back to exactly this kind of symbiosis we've been talking about, right? The place that AI can be really helpful is in taking tasks and busy work that doesn't necessarily need to be done by a human off of our plates so that we can do human interacting together more, spend more time together, and also so that we can expand the definition of what it means to do human work.
35:46 There's a lot of human activities that would be beneficial for more people to spend time on outside of the things we just talked about. Something I feel really passionate about is spending time together in community. Compared to 50 or certainly 100 years ago, we spend so much less time in our physical communities together. Whether you live in a city or in a rural area or a suburb, maybe you know your neighbors, but people are much less inclined today to be part of something they would identify as a community compared to 50 or 100 years ago.
36:18 And something I'm really curious about is how can AI be part of the story for helping to rebuild some of the community spirit that is so important around the world? We have a program called Claude Corps where we actually train people – especially young people who are coming out of graduate school, undergraduate school, trade schools, high school, really every different level of education – to say, how do you leverage AI to make your community better? How can you leverage AI, how can we teach you and equip and empower you with the skills to be entrepreneurial and start a business that's great for your community or support a soccer league or become a teacher? The more that people, particularly young people, feel like AI is empowering them to live the types of lives they want to live, I think that's us achieving our mission. My hope is that in the future we'll see more programs like this where we're actually helping people to learn how these tools can support everybody in getting to the type of world they want to be living in. That's really important because giving and service are not as central in most people's lives now, and yet they can add so much to becoming this best version of ourselves. You came to your current job in a very untraditional way. And here we are at this health conference with Alice Walton, who also came to this in a very unexpected way, because this is a medical school here where humanities and the arts are at the heart of training doctors. So what does this background, in your case, help you see about AI and health that others might miss?
38:15 Outside of the kind of humanities connection, I also spent the first years of my career working in global health. One of the things that I very acutely saw was this delta between health care in the United States, which certainly has a lot of room to improve as well, and health care overseas, particularly in the Global South. And something that really struck me was just the level of access to basic tools was so much lower where I lived in East Africa compared to in the United States.
38:49 And something that has been really powerful and inspiring to me, having done this work, is seeing the ways that AI is transforming health care access, opportunity tracking, data, infrastructure, information, everything everywhere around the globe. And outside of some of the stories we've already talked about, some of these cases are just really built for artificial intelligence to be able to help. An example that I'll give you is working with a partner organization. One of the things that we've seen is Claude's ability to access and interpret data that already exists can have a meaningful impact in the future in helping people achieve better health outcomes.
39:36 With IRC we've seen data around how many pregnant women are getting treated in a clinic before they give birth. The number we used in the study was about 6000 women that were tracked. Then at the moment they're giving birth, or the period of time that they're giving birth, how many of them are in a medical facility or are with a medical provider? And what IRC found was that was about 1700 women out of the 6000. So just having the data to know, wow, 3300 women are giving birth without the assistance of a medical practitioner — that's going to be much worse outcomes for those women than ones who are in a hospital or with a doctor — enables some of these partner organizations to be able to say, wow, in the future, when we are tracking the next cohort of pregnant women in this country, how do we make sure that at the moment they're actually giving birth that someone is there with them? That's not actually a medical intervention, that's just taking data that already exists and making it accessible and actionable to the people that are using it. And so there's so much low hanging fruit that Claude can help with to give physicians and medical practitioners and governments information to be able to say, here are some of the challenges so that we have the tools to even be able to come up with solutions. And what about an everyday case? You know millions of people talk to Claude and talk to AI generally about symptoms, medications and emotional struggles. At what point do you think a responsible AI would say, “I don't know” or send the person to a human? And how do you keep it from telling someone things they just want to hear, this flattery problem? This question about sycophancy is extremely important. We have an entire team at Anthropic that just works on this problem, which is the incentives that you set up, if you want the model to always agree with the person asking, are quite bad. And so we have done a lot of work to reduce sycophancy in our models because of many of the reasons that you suggest, and also because it just feels disingenuous, right? If a model just says, “you're right, Daniela, you're right. You're so smart. You're right about everything. You're right in that conflict.” Is that promoting the level of self-reflection, curiosity, learning, growth, humanism, that we actually want to see in the world? And my answer is no. Sometimes you need someone to say, actually, you might be wrong about this, or here's a person you should go talk to.
42:22 And Claude is actually quite sensitive to pushing the person interacting with them to other people. So Claude tends to say, hey, email your doctor with this question, or let me help you figure out the right questions to ask your doctor. Because, again, Claude is a powerful partner, but you can only be diagnosed through a clinician. Claude is not a replacement for medical care. You ask about individual cases. I will share my individual story for me and how Claude helped me in a medical situation. When I was seven months pregnant with my daughter, who's now one, I got a really bad bacterial infection, but I was misdiagnosed by three sets of doctors — by my OB, by my GP, and in the emergency room.
43:06 And this is not because any of those are bad doctors, right? They're really busy. They're looking at a roster of symptoms. They have a big patient caseload. But the result of it was that I was in a very serious medical situation that could have led to a really terrible outcome for me or for baby or for both. And Claude was actually able to properly diagnose me and say, actually treating a bacterial infection isn't that hard. You just take antibiotics. But because I hadn't gotten the proper diagnosis, I was in a pretty bad spot. And so the ability for me to say, hey, I'm going to go to my doctor, it was actually my husband who at that point had to go to my doctor and say, hey, is it possible, could we test for a bacterial infection? Just the ability to have the knowledge as a patient to say, could you test me for this ultimately resulted in an outcome that was much more favorable for me and baby. I think about if I hadn't had that intervention, what might have happened? Unfortunately, it could have been a really different ending to that story.
44:08 And we hear stories like this one every day. People email us saying, this is what Claude did for me. We have a user happiness channel where just employees share stories about how Claude helped them or helped a family member. It's really quite moving, the degree to which I think it is providing real, tangible benefits to humans today to get better access to care. I love this story as another example of a symbiosis in which your husband got that information, took it to a doctor, and that led to an incredibly positive result where you now have your one-year-old. And it's really also part of Alice's vision, which is that health is complicated. It's physical, it's mental, it's social, it’s spiritual — it's everything together. And the other thing that I'm very passionate about is that we now have so much science that makes it clear that health is also what happens between doctor visits — like how we sleep, what we eat, how we move and exercise, how we manage stress, how we connect with ourselves and others. These five daily habits are instrumental in our health, and I'm really excited about how AI, through its power of hyper-personalization, can come up with the right nudge, the right recommendation, to help us every day improve these daily habits. And is that something that you think should be given more attention?
45:54 I have certainly personally experienced this, where sometimes, even if I'm talking to Claude, I'm asking a question, “should I call my pediatrician?” or “should I be worried about this symptom in myself?” Its ability to pull information from prior conversations to say, “this reminds me of when you had this other symptom a year ago.” I think there's something very special and unique, in concert with a human doctor, about the ability to have this tool that is keeping track. It's the data fact finding problem where all the data is there but it's really hard for a human, particularly a doctor who might have hundreds of people that they take care of, to remember all of those facts about you. They're going to look at your chart, but your chart is maybe going to have your allergies or any medication you're on, it's not necessarily going to have this full picture of you as a human from every angle. And I think there's something potentially really powerful and self empowering about humans to be able to say, wow, this is this tool that can help me understand my own health, wellness, and happiness better. How can I then take that information and use it to partner with my doctor, or with myself, or with whoever, to really live a healthier, happier life? I think it's that ability to have that context, right? Humans are very complicated. And so your doctor kind of sees one angle of you but it's hard to remember or recall the level of information that these models are able to synthesize from multiple conversations with a person. So now that we're at this incredible moment when AI is, or if not right now, is soon going to be more intelligent than we are.
47:51 How can we use this moment for humanity? Can we just say, okay, AI, you own intelligence? Can humans own wisdom? Homo sapiens means wise man and wise woman. It doesn't mean intelligent man. So if you think of it, since the Enlightenment and the Industrial Revolution, we started defining ourselves by our intelligence. I think therefore I am. But our wisdom, our soul, our spirit, our consciousness, whatever name you want to give it, is so infinitely more important and more powerful. So can we actually see this moment of AI dethroning intelligence as our primary identification as an incredible moment where we can say, great, let's go beyond intelligence?
48:55 And I feel women here have an incredible role to play, because traditionally, women — going back to Athena, the goddess of wisdom, to go back to Greek mythology — have had this connection with the deepest parts of our humanity, partly by giving birth, partly by being so rooted in the community. So what do you think if we let Claude have intelligence and humans have wisdom?
49:30 Yes. I think so much of the interesting core of what I hear from you is this concept that intelligence is just one facet of what it means to be human. And we use our intelligence for a lot of different things. We use it for medicine, for science, for literature, for curiosity, for relationship building — for so many different things that I would call wisdom. And so really what we're talking about is this one branch of it that AI I think is very powerful at, which is the ability to compute things, to do mathematics and to synthesize data at a superhuman level. But there are so many other things that are on the spokes of that wheel that you're talking about that I find deeply human and that I don't think AI can replace.
50:21 I always come back to this principle that fundamentally, humans value spending time with other humans, and humans value self-reflection and creation and curiosity. To go back to the symphony example, I do not think that in a world of AI people will be going to the symphony less. I think they'll be going to the symphony more. Right? Because AI may be capable of producing text and music and art and all of these things, but people will not want to consume it the same way that they consume a symphony performance where humans are actually performing on a stage for them. Claude can never do that. Claude can never be your manager or your therapist or your partner or your kids. Those are fundamental human relationships that I actually think will become more important, we will spend more time on, and we will value more in a future of artificial intelligence than we value it today. This is a great moment to end and just go through a very quick lightning round. Great.
51:28 So what keeps you up at night besides your children? I think, other than my kids, worrying that artificial intelligence will become such a politicized topic that it will be hard for us as a broader society to have the type of important conversation that you and I are having right now. Like I've said a few times, this is going to go better if we all feel like this is something we're doing together versus it happening to us. And if this becomes overly politicized, it's just going to be hard to have the level of nuanced conversation that you and I have had today. Love that. What’s one question you hope we’ll be asking about AI one year from now? I hope and I expect that we'll have this conversation again a year from now and many of the questions we talk about will be, tell me about the scale of impact that AI is having to produce positive benefits and change the world for the better today. What we're seeing now is that these beneficial deployments around global health, around education, around nutrition, around biology and life sciences, is it feels like we're at the base of this exponential curve, and we're starting to really see it take off in terms of what artificial intelligence can do for biomedical research, for science, for global health, for education and for all of these really important areas. I think a year from now we're going to be measuring many of those impacts on a much larger scale. And so hopefully it will feel like 10x when we talk next. I love that, we're in the first innings, right?
53:11 So what does your five-year-old think that you do all day at work? So a funny story about this: he is now deceased, but there was a wonderful alligator, a white albino alligator, named Claude at the California Academy of Sciences. And one day, for his birthday, I actually took my son to the Academy of Sciences. He was three. So this was two years ago. And he said, when we were looking around, “Mama, Claude!” He hears me say Claude at home a lot and so I'm pretty sure he still thinks that I'm a zookeeper or an animal keeper at the Academy of Sciences. I have not done that much to dispel him of that image. Even though poor Albino Claude has since passed away. We have to find another Claude alligator. So who is one person who has inspired you and your work or your life?
54:03 I hope this won't come across as corny, but Dario is a person who has inspired me a lot in my life. He worked in biophysics, he worked in biology, he did early cancer research, he got a PhD in physics. That was his passion. That's what he loved to do. But he really pivoted his postdoc to have more of an applied biology and science component to it because he really wanted to help save people's lives and help make people's lives better.
54:31 And getting to see the biomedical and science related applications of AI today, it's so wonderful to be able to watch him kind of come full circle and work on that. And I think he feels the same way and I like to think he feels the same way about me in global health. I started my career in global health, really wanting to have an impact to help people everywhere in the world have access to basic health care, be able to live healthier, happier, longer lives for them and their children.
54:58 Our ability to make that better and influence that via the tools we're building today is such an incredible part of my job now. And so I think just both of us being able to watch this old part of our careers kind of come to fruition at Anthropic has been really powerful. It's an incredible relationship and how you complement each other. So I know you like physical books. What books do you have on your nightstand? I have to think about this. I actually have four right now. I'll go from the most literary to the least literary. The first — well, these two are maybe tied — is a collection of short stories by E.M. Forster. Have you read The Machine Stops?
55:42 Yes. An excellent, excellent short story. And E.M. Forster has beautiful prose. A friend actually gifted it to me recently for my birthday and I read it a long time ago. But I read it again because I think this concept we've been talking about — sort of physical and social disintermediation — it felt quite prophetic from that. And some of the themes we've talked about – human relations and people being together, and love of nature and the human spirit – feels very powerful in that short story in particular.
56:16 The second is Leaves of Grass. I love Walt Whitman. I've read it many times, but I have it on my bedside table because sometimes I just like to read. I should say I work through all of these books very slowly. The third is — I'm getting now into the less literary — there's a book I'm reading now called The House in the Cerulean Sea. And it's sort of science fiction fantasy. It’s about — it's a very charming book, I'm only about a tenth of the way into it — young magical children in an orphanage. It's very interesting and lovely.
56:55 And then the least literary, I am reading Yesteryear, so I have to disclose that. That's great. I think it's amazing that you have any time to read. I love that. Very slowly. If you were the world's AI czar, and I'm very happy to nominate you, what would be the first thing you would do? I would probably start by saying I don't think there should be a world AI czar. I think it's too — Just a czarina maybe?
57:26 It's too significant of a technology for it to be governed by one person or for one person to be responsible for it. So I personally think finding a way to make this more of a conversation that many people feel the ability to participate in and have agency over feels incredibly important to me. So as czar I would probably abolish the concept of me being an AI czar and can we find a way for more people to be included and empowered in the conversation of what should happen with artificial intelligence? Thank you so much for this incredibly inspiring conversation. It was such a pleasure. Thank you so much.