The power of AI, from curing cancer to ballot boxes

0:00I am so excited because we have a special Read Riffs today. at the top of our episode, we're welcoming a past possible guest who is an acclaimed author and company founder, Siddhartha Mukherjee. Sid, it is so great to have you on. Given an amazing second launch of one of my favorite books of all times, Emperor of All Maladies, Ari and I have decided to invite Siddhartha Mukherjee, who is both the author of said book and also an amazing oncology researcher and a CEO and co-founder of Manus AI, which we're doing together. We thought that since the second edition came out of this book that I first met you with, Sid, this is like my very first encounter of Sid was reading this book and going, oh my God, this is amazing.

0:45So Sid, say a little bit about how your path into becoming an oncology researcher and then writing this Pulitzer Prize winning book. I started off in medical research. I went to Oxford and then became more and more interested in cancer. I was first interested in viral cancers. By viral cancers, I mean cancers caused by viruses in humans. Cervical cancer is a good example. And then became more and more interested in cancer in general, decided to get a medical degree and became an oncologist. I thought it was the most interesting problem that anyone could hope to solve. The book was a journal. It was a journal that I kept when I was training in oncology.

1:22And I was a trainee, so you have to realize that I was at the bottom of the flagpole. I began the journal because a patient said to me, she was getting treated for lung cancer, and she said to me, how did we get here? And I thought to myself, you know, actually, even I don't know how we got here. I'm not sure how we got here today. Why are we using this treatment? How do we think about cancer in this way? And if I don't know, what about the hundreds of thousands of patients and their families and loved ones who encounter cancer? How did they get there? And so that's how I started writing the book.

1:57As I said, it was a journal at first and grew and grew and grew. And then it became what my mother called the telephone directory from hell. There were so many pages bound in black. And so that is how the book was born. And so why a second edition now? And what are the updates for the book? 15 years have passed. And in those 15 years, enormous things have changed in the landscape of cancer. biology, cancer research, cancer medicine. And yet cancer remains the meme for everything that is malignant. And yet people use the words still to signify so much metaphorically. And so in those 15 years, you know, I asked myself the question, how much has changed?

2:40What has changed? And I really felt as if the book needed an update because science moves so fast. It's very interesting to write a book 15 years ago and then realize that you have to write another book 15 years later. People generally don't bring out completely new second editions of their book, but it felt very necessary. And so that was that sort of the purpose. To answer your second question, what's changed? So very broadly, I would say that we think about cancer in four areas, I would say. Number one is prevention, cancer prevention. What's changed in cancer prevention? Well, we are discovering at last, at long last, a whole new universe of potentially carcinogenic agents that don't act in the way that we thought that carcinogens, as in cancer-causing chemicals or cancer-causing agents, usually act.

3:25In other words, they don't cause mutations in cells, but rather they create an environment for cells to be able to divide, to form tumors and become malignant. So this is a whole new class, and I think we will discover many more. A great example of this, by the way, is air pollution, which we never knew. Another great example which is emerging are potentially new chemicals that we've invented. There's a whole debate going around whether these forever chemicals might be implicated. So that's one arena. Very, very important, very exciting, new research, entirely new direction for cancer. Okay. Second is what we call early detection.

4:06Can we find cancer early? And if so, what do we do about it? So once again, a huge difference. So we've heard about blood tests that pick up early cancers. So there's a whole chapter on those. People are going through full body scans. Are they worth it? Are they picking up something that we didn't know before? What do we do about them? So there's a chapter on early detection. We weren't doing that, obviously, 15 or 20 years ago. And finally, a chapter on treatment. Major theme is immunological therapy and immunotherapy, which is curing cancers such as melanoma, skin cancers previously considered incurable.

4:45T-cells or T-cells that have been engineered curing cancers such as lymphomas, leukemias, myeloma, previously thought incurable. The whole book is capped off by what I call a new vision or a physiology of cancer. What happens next? How are we reimagining cancer? Because it's all about imagination and about experimentation and, of course, science and medicine. So, Sid, I love that your book, it traces the progress against cancer and how it comes in waves. Like there is basic biology breakthroughs, therapeutic revolutions, there's political mobilization, of course. And it's been super fun to have sort of a front row seat at everything you all are doing at Manus.

5:24Do you think the reason for Manus is that we're at sort of the forefront of the next wave of cancer research and treatment? So that's an excellent question, Aria. Reid, Ujwal, and I co-founded Manus because we started firmly believing that information revolution, every information revolution has had enormous effect on medicine. Think about the basic computer revolution and its effect on how medicine was and is practiced in terms of how we keep notes and how we can compare notes and switch notes. think about the information revolution in digital technology, imaging, and CAT scans and MRIs, it seemed as if, well, it doesn't seem as if it is becoming amply true that the AI revolution, which is, among other things, a data revolution, among other things, a computational revolution, a learning revolution, it became amply clear that if medicine, especially cancer medicine, wasn't going to embrace it and move forward with it, that we would lose momentum.

6:30And that's how Manus AI was formed. Manus AI is an AI-native biopharmaceutical company. One of its main focuses is using AI as a device to understand how to make medicines. Not AI as a data management engine, not AI as a clinical trial engine, but using basic AI and physics and chemistry to find how one drug, a new drug, how the new chemical matter can interfere with the growth of cancer cells. So for the layperson who's watching this space and might not sort of have the deepest understanding of drug discovery or how that interplays with AI, and Reid, feel free to jump in on this one. Sid, what is the difference?

7:12We're seeing lots of startups claiming that they're using sort of AI plus bio to cure cancer and other diseases. What makes Manus different in the way you guys are approaching this? Well, we are approaching it in two fundamentally different ways. One of the ways that we're different is that we're not a software company. We make medicines. One of the first questions that Reid, as an entrepreneur, asks his fellow entrepreneurs is, what does your company do? And it's amazing that you can ask this question to many, many, many companies and not get a straight answer. But what does Manus do? Manus makes medicines.

7:49How does Manus make medicines? MANUS is building a foundation model, what we call a molecular oracle, which enables us to, at an atomic level, with atomic level precision, understand the dynamics of a cancer target, one of those proteins that is driving the malignant growth of a cancer cell. And again, with atomic-level precision, place atoms inside that target that will block the activity of that target. It's a little bit like how you might place go pieces. In fact, one of the modules we call the molecular go. It's a little bit like how you might place go pieces on, in this case, a three-dimensional go board and figure out where the optimal pieces would land such that this protein can no longer activate all the cascades it activates in order to make a cancer cell malignant.

8:53And Reid, you obviously are using AI in so many different ways to attack so many different problems. What is the reason that you thought AI plus cancer would be so powerful? You know, building upon what Sid said is that I think the best way for AI to actually make a huge difference in medicines is to recreate the medicine discovery creation process, being AI native, starting with AI in a set of things you're doing. Now, most Silicon Valley people says, well, that means doing everything in Silicon. No, you're doing a lot in Silicon, but not everything. There's a lot of reasons not to do everything in silicon.

9:36And there's a set of other different kind of things that, as we make medicines, will be more revelatory of. But a simple way of looking at it is when you saw the current AI revolution, which ended up using scale compute and scale data in order to make predictions, for example, predictions that led to Move 37 in Go, So predictions that lead to being able to write an essay of an original sort with a large-scale LLM predicting the next token as a way of doing this. The question around making medicines is the prediction of which molecules might be the right therapeutic for a particular disease, a particular cancer.

10:34And that was kind of the, oh, this is why I think AI could be really stunning if done in an interesting new way. And the way that Manus came about is I had known Sid as an amazing author. I'd known him as a world-class cancer researcher. I'd known him as a kind of a general intellectual. And I was like, well, I have this idea about AI. I'll go chat with him. And he's like, and by the way, not only is that an amazing idea, but cancer is the right subject because it basically, it kills all ages, all races, all genders, et cetera, et cetera, et cetera. And it's just like, this is the thing for us to focus on.

11:24Now, there's obviously a bunch of other things that become possible, abandoned diseases, and a set of other things that are also very important for human health and humanity. And then the thing I learned about Sid and this amazing, Sid also has a minor superpower and choosing amazing restaurants, in this amazing Japanese restaurant that we were talking about, he's like, and you know I brought drugs to market. And I was like, no, I actually didn't know that. Because the other things shone so bright, I couldn't see the history of entrepreneurship in the past. And so this is what led to, and we should do Manus.

12:06And along with Ujwal Singh, we're anticipating 2026 to be a really great year. Sid, thank you so much for joining us. I'm so excited to read the new edition, which for everyone listening, it is out now. So please go grab a copy and also, of course, follow all of Manis' work. And for those who want more from Sid, we actually did a full episode with him two years ago. So look for it in the feed and tell us what you think. This podcast is sponsored by Google. Hey folks, I'm Amar, product and design lead at Google DeepMind. We just launched a revamped vibe coding experience in AI Studio that lets you mix and match AI capabilities to turn your ideas into reality faster than ever.

12:47Just describe your app and Gemini will automatically wire up the right models and APIs for you. And if you need a spark, hit I'm feeling lucky and we'll help you get started. Head to ai.studio slash build to create your first app. Loved hearing from Sid. And now we are going to turn our attention to some Silicon Valley tweets that have been happening in the last few days. A few days ago, Patrick Collison tweeted out something about how the Valley is ripe for new frameworks. And I want to get your thoughts on it. So I'm going to commit the cardinal sin of reading a tweet on a podcast. But here we go, just in case anyone didn't hear it.

13:22Patrick Collison said that two conversations this weekend made me think that there's a vibe shift afoot in Silicon Valley around what one should work on and what is worthwhile. Culturally, it feels like the moment is ripe for new frameworks. Davos expert morality is stale and discredited. It's also apparent that the just-be-super-based counter-enlightenment is not really an answer. Yes, woke went too far, but simply inverting it doesn't work. EA is no longer the automatic default for smart people. And there's increasing skepticism of slot and slop machine dynamics. So overall, what is worthy and valuable feels like it's becoming more central.

13:58So I actually loved this tweet from Patrick, especially because I think a lot of people in the Valley listen to him and really care what he has to say about what it is valuable to work on. So I would love to get your thoughts. How did you take this tweet from Patrick? Well, as usual, Patrick is extremely thoughtful. I mean, one of the things that I like about Patrick is not only is he a kind of amazing and smart entrepreneur, but also thinks more systematically, you know, whether it's the acceleration of science, whether it's the pattern of entrepreneurship, whether it's kind of what's going on within, you know, kind of patterns on the internet, culture, et cetera.

14:38And so, you know, the fact that he is, you know, not just kind of, you know, creating straight press, but also kind of commenting on kind of what is the culture of the valley? How does that target, you know, kind of things that matter, Things that matter in entrepreneurship, things that matter in business, things that matter in industry, things that matter in society, things that matter in humanity, things that matter to science, et cetera, as ways of doing it is exactly the kind of thing that I think makes him one of the important leaders, both in the Valley and the world. And I think that the question around kind of saying, hey, what actually really matters is building important technology that makes a big difference for industry, for society, for humanity is the thing that really matters.

15:34And I think that the – this is obviously one of the reasons why I've been giving these speeches in Italy, Bologna, Perugia, for remembering what it is that was why the Renaissance actually in fact really matters. Not just because it was an enormously generative period that all of a sudden we moved to an acceleration of science, acceleration of art, acceleration of humanism. But that we also were on what kind of things really matter. in terms of this. And that's, you know, to have this kind of popular trope in Silicon Valley about how Silicon Valley can be, you know, the next Renaissance, the next Florence, the next, you know, Bologna, the next Perugia, you know, as, you know, part of what's being created, it also should be embodying this what matters.

16:25And what matters, of course, is the elevation in humanity is the kind of the theory of human nature, the theory of what journey are we on? What can we become? How should we do that as a part of it? And, you know, I think part of it, you know, the classic, you know, Patrick is stating this in kind of a classic Silicon Valley way, which is, you know, down with the old, the complete invention of the new. I think it's actually important to be learning things. Like actually, in fact, I think there was a bunch of stuff that was important in the Davos Assembly in terms of, you know, like, okay, what are all the voices across the world?

17:07Global South is part of when doing this and being in dialogue. And I think that's, you know, while there's like the bureaucratic, you know, nonsense of that's not how the invention of new science, that's how the invention of new technology happens. But, you know, The dialogue is important. And then similarly, like, for example, to take EA, I think EA has a bunch of challenges to it. But the notion of saying, hey, let's treat every human life as roughly equivalent and be maximizing the quality days of human life when you kind of think about it across the globe is not something that's outmoded. That is, I think, actually still an important thing.

17:47Now, the way to think about that, the way to get there, I think, is important. And I think that there's – anytime that something starts becoming a religious canon – and by the way, similarly, like the, hey, let's just be bold and genuine and totally raw on the being quote-unquote super-based, those are good elements for entrepreneurs. Those are good elements for being contrarian and right. But I think that the question is to actually think seriously what you're doing. And my gesture overall would be a theory of humanity, a theory of human nature, a theory of how technology evolves that and being something better than the HBO show in terms of how you're articulating it.

18:34And that's part of the reason why I think Patrick's call to arms or perhaps call to meaningful technology is such an important call. And do you think sort of those ideals are right? First, don't tear down the past. Let's actually take the best parts of these waves that have come previously. But do you think the valley is heading in the right direction towards more meaningful work? And are there things you would say to entrepreneurs who are just starting out now that sort of nudge them in that right direction? It can be. It's one of the reasons why I'm such a strong advocate of Silicon Valley, even amongst its various absurdities.

19:15It's a little bit like you say, hey, what's the range of crypto? Well, obviously, there's a lot of absurd stuff in crypto. But on the other hand, of course, crypto has the amazing elements of the next generation of the internet technology in terms of identity systems, value systems, banking, recreation of financial systems. It allows the high-functioning financial systems around the entire world. And those are kind of the things that crypto has the trajectory to potentially create in of much as also like you got to get rid of the garbage, minimize it, can't get fully rid of it, and increase the amazing transformation of the world.

20:01I think it's kind of a similar thing where I think at the moment, the Valley, it's all obviously very focused on AI. But I think that the question about how it manifests itself is, I think, still in its earliest phases. And I think that the call to how do you work on this in very meaningful ways is exactly the right kinds of questions that people should be asking. That's what I hope for. So, Reid, you mentioned the speech that you gave in Bologna, which I was lucky enough to attend. And I think, again, you were talking about the similarities between the Renaissance and sort of Italian history and what could be and what has been in Silicon Valley.

20:40Can you say more about that? Bologna was the first of these Italian speeches, which was a great honor to give to the Bologna Business School and the folks there. One of the things when I was thinking about it is that the progress of humanity comes through this kind of Hegelian dialectic Thesis, antithesis, synthesis Which is you start with a thought, you go to the opposite, the contradiction of it And then you try to build, as it were, the best of both worlds The collection from what was the original thesis that was interesting the antithesis that challenged it, and then the refinement and the synthesis.

21:22And then this, iteratively, is part of how we make progress. We can make progress scientifically, we can make progress in society, we can make progress in governance systems, and so forth. And I think that the important part of the work is to see what's good in both the thesis and the antithesis to draw it into a synthesis when you're making progress with this. So historically, part of the thesis here is to say, hey, we're building infrastructure that kind of rebuilds society. Or for example, in the case of Davos or political discussion, we're having a discussion about what really matters in the world and what kinds of agreements can we get to and what kinds of The antithesis is to say, okay, well, you actually need a creative destruction where the exact patterns of this industry, the exact patterns, it needs new technology, it needs a new way of working.

22:32and then that, you know, you're creating alternative technology. And the synthesis should be something that's along the lines of a, call it a, you know, what's our theory about how this is good for humanity overall? And this is, of course, one of the reasons why at the beginning of this year, you know, Greg and I wrote Super Agency because it's kind of a focus on, you know, not just what is great technology, but what gives people more agency? What gives them more of a sense? And by the way, it doesn't mean all of your previous agency is preserved, but the new agency should be kind of a collective set of superpowers, which then enables our humanity in much better ways.

23:16And that's, of course, what we think is the right lens for thinking about what kinds of AI we should be putting our efforts into creating. Well, speaking of what kinds of AI we should be creating, earlier this year, you and I actually talked about how excited we were when we heard about some AI persuasion tools that was working with people who believed in conspiracy theories. So AI was very effective at getting flat earthers to rethink their views, for getting people to believe in the moon landing, some of these sort of outrageous conspiracy theories that usually when you say like, I try to convince someone with facts and it doesn't work.

23:53But actually, a chatbot can be infinitely patient and they can sort of tailor what they're saying to help these people come to come to understand actually facts of the case. And so that is a side where persuasion can be used for good. Recently, we saw two new studies come out in Nature and Science, and they found that even brief interactions with AI chatbots can meaningfully shift people's political opinions. And they found that in some cases, chatbots were about four times more persuasive than U.S. television campaigns. And between 4 % and 10 % of voters in some counties and countries switched to supporting the chatbot-backed candidate after just a single conversation.

24:31And so are we entering a zone where we're going to, instead of political ads all over the place, we're actually going to have chatbot persuasion wars? And it might depend what chatbot you're talking to could sort of meaningfully sway who you're going to vote for on Election Day. So should we – what should we do about this? Is this something – how should we think about it? Should there be government intervention? And how will this affect the political climate? Well, I think not surprising since you know me pretty well. the way that I think that we should think about the beginning of interventions is measurement and transparency.

25:08So the kind of a question around, you know, like, what are the things that, you know, should be, you know, advertised about like, okay, is this chatbot, you know, have a position, you know, so for example, if it's a, you know, if it's Grok, it should say, well, I've got all these meta prompts about, you know, how to say that Elon is, you know, the best athlete ever and the most intelligent scientist in the world. And, you know, dah, dah, dah, because, you know, I've been trained to be the court gesture of saying, you know, how amazing Elon is. And so, you know, but whatever the specific thing that you're doing for a chatbot, you should, you know, have your kind of principles, how it's designed, kind of what you're accountable to in it.

25:55And I think that kind of thing of statement about what your kind of design and so forth is and what your efforts are for it. And obviously, most of the quality chatbots will do things like saying, hey, we're trying to be reflective of the best of current scientific knowledge. We're trying to be, you know, kind of, you know, answer what you try. But we also try to avoid conspiracy theories and we try to do other things as ways of doing it. And then, you know, obviously being able to have a dialogue by which it's specific, not, in this case, ad hominem slogans, which is the most common thing. Like, oh, it's a woke chatbot.

26:38And you're like, it's a woke chatbot because it's, you know, kind of willing to be critical of President Trump or it's a woke chatbot because it's not giving you a, you know, a riff on Biden, you know, et cetera, et cetera. is like, okay, these are kind of silly arguments, but credible people make them, which is part of what makes them so absurd. And so I think that's the kind of thing that's going to be important. Now, the short answer is we're going to have lots of failure points. We're going to, you know, in this politicization of the information age where, you know, it's so badly politicized that, you know, you can't have a coherent conversation around something that's relatively scientifically understood like vaccines.

27:26That's like, oh, that's political information. And it's like, no, actually, medical studies are not political information. Doesn't mean they're always right. Doesn't mean that there isn't an evolution over time. Doesn't mean that, for example, they say, well, but people are saying that vaccines never have any side effects. Well, only idiots say that. It's like, that's not an argument. Anyway, so back to AI, We should want these things to be good foils and participants in truthful discussion. Now, part of the problem, of course, is we as human beings have to get to truthful discussion. And right now, I'm not enormously optimistic about how well we're doing the truth discussion.

28:06So, Reid, if you were running a frontier model company, if you were running a chatbot, you were the CEO, would you let it answer the question, who should I vote for in the next election? Do you think that's OK? I think I would let it answer the question, but I think the way that I would answer the question, like I wouldn't say, you know, vote for Sarah or vote for Bob. I would say, well, the theory of voting for Sarah is da-da-da-da-da. The theory of voting for Bob is da-da-da-da-da. And here is what, you know, some – like if you cared about, for example, economic prosperity, here is why tariffs are actually a terrible idea.

28:46Right? And actually don't work. And, you know, and they say, well, but it hasn't been working. It hasn't been doing that badly so far. It's like, well, we're actually the tariffs are decreasing the major economic boom that we're getting from the market with the AI investment already. And so it's like it's actually decreasing what the benefit of the last 10 years of building towards AI prosperity has been. And so it is actually, in fact, a decrement, but it's a decrement in a wave that is overall propping up economic stuff. But that's, of course, part of the reason why people say, well, you say the economy is doing better, but prices are going up and affordability is a challenge.