AI, Personalized Medicine, and Cancer - Dr. Emilia Javorsky

Hi everybody, and welcome to this week's Talking Biotech podcast. Now, the promises of AI have been many, and we're seeing it incorporated into more parts of our lives. It's a useful tool. It's a way that helps us get things done sometimes. But there's also a lot of promises that have been made that when this comes, we're going to solve some of the world's most pressing issues, one of them being cancer. And we've seen a lot of discussion about how AI and new AI tools are going to help us break some of these challenges with some of the most insidious diseases known to humans. The problem is, does it overstep? And sometimes we have to rethink and recalibrate our concepts and our rethinking about what are the limitations and what are the ways that maybe AI could be used in ways that we previously haven't used it in future applications. What can we do differently? And with that, we're speaking with Dr. Amelia Jaworskyj. She's the Director of Futures at the Future of Life Institute in Boston. So welcome to the podcast, Dr. Jaworskyj.

Thank you so much for having me.

Yeah, my pleasure. I think this is fun because we all can use a little AI sobering, but not just dialing it back, but dialing it forward in the right way. You've had more sophisticated thinking on this than anybody I've talked to. So let's dive in there. First of all, you know, you're an MD by training, or tell me more about your background before we dive in.

Yeah, so by background, I am an MD by training, and before that had a previous life in public health, and I've worn a lot of hats. across the biomedical discovery spectrum, from being in the clinic to being a postdoc in a lab to founding a couple of companies and still kind of keeping an academic hat as a biology researcher in addition to what I'm doing now at the Future of Life Institute. So I've seen it from early discovery at the bench kind of all the way through to the bedside.

Very good. So let's start out at the beginning. And when we're talking about cancer, I've always objected to the idea that we're going to cure cancer because is that you can't, it's not one thing, right? So maybe let's start out by defining what is cancer?

Yeah, that is a great place to start because I think cancer, like AI, is somewhat treated as a monolith in a lot of discussion, like it is one thing when it is actually probably one of the most complex diseases, if not the most complex disease that is known to humanity. And I think we can see this in the evolution of how we've thought about cancer in the scientific and medical community. So, in the earliest days, cancer was thought to be one cell gone rogue that divides a lot. And if we can just stop those dividing cells, then poof, we've cured cancer. And as we've unlocked more and more tools to measure and understand biology over the decades, the thing we found is cancer becomes more and more complex and more and more complicated and more and more nuanced, not more simple. And so it's evolved from being thought of as a disease of cells gone rogue to a really complex disease process involving the immune system, involving blood vessels, involving the cells themselves, involving our microbiome, right? And so, as we've undertaken our study of cancer, it's something that becomes more complex and more individual, not less complex and more generalized.

And at the same time, our understanding has become more complex. It seems like the promises have become more simple, especially as they're amplified through traditional and social media. So what are some of the ways that you've noticed that AI is poised to cure cancer?

Yeah, so I think this AI cure cancer promise kind of obscures a lot of the great work that's actually happening in leveraging AI tools that we already have around today that scientists academic labs and pharmaceutical companies are already using to make progress. So we've seen this from AI helping to design new molecules, AI helping to better predict whether that molecule might be toxic or not. So doing some better predictive safety testing in the lab and helping us to better design clinical trials, to help design sort of what we call synthetic control groups, which basically is like instead of having just someone that gets disease in a control group, can we simulate that so we don't have to basically waste a patient's experience on something where we already know the trajectory of the illness? We've seen AI being used in the operating room to help us define margins to know if we're clear or not when removing tumors, right? So there are many ways that AI is already being used in the clinic and in the research process that I think get overshadowed by this bigger promise that AI is going to cure everything. Because there are many ways it is actually already moving the needle that don't wait on sort of future AIs as panaceas as they're often presented.

Is the word cure alone a problem? Because we seem to throw that one around a lot. And I know physicians and researchers are hesitant to use that. But should we really focus on a different metric, something like maybe management or other ways that we can So, what is the difference between resolve the problem that isn't necessarily cure?

Yeah, I think the word cure is problematic for most complex diseases. Like, we've yet to actually cure a complex chronic disease in humans, right? We can cure things like infectious diseases and make them go away forever, but many other diseases, management, I think, as you suggest, is the right term to think about these things. You can manage diabetes, you can manage high blood pressure, and now, as we're seeing with cancer, because you are are using so many different modalities to kind of lean into the complexity of cancer, it becomes a disease management strategy that you have for patients. So patients that do go into remission, you know, we'd like to call that cured, but what they are is sort of turning cancer into what was a short-term death sentence to long-term management of a chronic disease. And I see that as kind of like the success state that we can most reasonably aspire to for cancer.

In your products, you've argued that raw intelligence is the primary hurdle in curing cancer or addressing cancer. And if a super intelligent AI were to emerge tomorrow, but was restricted to our current biological datasets, how much progress would we actually see?

Yeah. So I think this is one of the big missing ingredients in this AI promise for progress in biology and medicine in general, which is, well, what data is it actually going to use? Do we have high-quality clinical data of what health looks like, of what disease looks like at an individualized level, at a population level, to actually give to an AI to derive insights from? And so I would say if we had a super intelligent AI today, would it probably make some progress in areas that we have great data for? So we have great cell data, we have great preclinical data from mouse studies and things like that. So it might be able to make some progress at that first phase. But the big problem is we know that, you know, 90-plus percent of things that work in cells and mice don't actually move the needle in humans, right? And so in order to bridge that gap, you need that human data. Getting more intelligent in the first part is not going to help you if you can't actually connect that to clinically relevant outcomes.

Yeah, but isn't it even worse where, you know, mice aren't people, but people aren't people? I mean, we have separate populations that would have different sensitivities possibly to a given potential treatment, certain therapy. And how much does that overlay cloud the problem?

Yeah, so I think as medicine has evolved, similar to the cancer story getting more individualized, we've seen this across different disease states. We now realize Alzheimer's disease is not just one thing. It's probably a set of many different diseases that present similarly with different root causes of biology in different people, right? So medicine as a field needs to kind of move away from this population level or this mean averaging to understand how can we lean into individual individualized biology and actually looking at what is happening in an individual and what biological processes are going wrong in that particular person. So in the case of something like AI, which is looking at general— to learn from general trends and looking for what's overrepresented in the data, that's kind of pushing you back towards the old way of thinking of like reversion to the mean. What's the average thing happening in the population? What's the most likely thing? As opposed to where we need to go in medicine, which is What are the more unlikely things in an individual that are actually driving the disease process in that person?

Yeah, well, let's take the same concept the other direction, and that, you know, tech geeks will often frame cancer as a code to be cracked, right? So as a physician scientist, what's the most dangerous misunderstanding from tech folks about the heterogeneity of cancer itself?

So I think that one of the big problems in the tech community, which is actually not new, we've seen tech try to come to medicine several times now, from the early AI drug discovery companies to other ventures into healthcare that have tended to fail for similar reasons, which is basically underestimating the complexity of both human biology and disease, and then also the healthcare delivery system in which those new therapies need to navigate in order to actually help patients. And so I think with the cancer piece specifically, it's similar to, I think, what people thought in the early days of the Human Genome Project, which was the idea of once we know our DNA sequences and our human sequence, we'll find the bugs and it's like code and we'll debug the code and then we're going to cure all these diseases. And people were really hopeful for that in the early days. There were projections that by now we would've already cured all of chronic disease. And I think the hope and thinking in AI is similar to those early missteps with the Human Genome Project, which is thinking that this is a simple code or a simple intelligence problem that can be solved and gravely underestimating the sort of foe that they're up against.

Yeah, I remember back in those days when we thought, oh, well, people would say this, right? And I'd always say, you know, I can give you the parts list of a 747, but, you know, you don't know how to make it fly.

Yeah.

And so this is the, this is really a very similar problem, but this is the problem with all the data is that, is it a question that we're looking at the wrong kind of data that we really don't even have the technology or the instruments to measure the dark matter of cancer biology? Like, wait, what? Is there something that we may be missing that— so you see what I'm saying? We're confined only to what we know. And so what are there things that we can learn going forward that will add to this?

Yeah, I think of that across 2 dimensions, which is like, how can we measure better? And then how can we measure more? So on the measure better side, it's what are new tools to interrogate biology and actually understand biology and move that needle and our ability to investigate at higher resolution and new resolutions. And we've seen this with the advent of sort of the multi-omics world. So we have the genomics thing, now people measuring the proteome, measuring the metabolome, measuring the immunome, right? And have all these sort of new tools coming online that we can measure human biology at resolution that previously was inaccessible to us. But then that brings us to the second point, which is how do we measure more, right? And most of those measurement modalities have just been limited to research settings or specific studies that have been done. They actually haven't been done at scale scale or integrated into routine practice. And until we measure a lot of people over time with that kind of high-quality GPS map of biology that we could generate, it'll be very difficult to actually translate those measurement tools into actual, you know, new therapies, new ideas without actually applying them at scale in people.

Yeah, and I think the evidence for that is pretty good because if we look at old datasets that are existing, so like something like the UK Biobank, that's been cited as being really helpful in shaping our understanding of a number of different diseases and conditions because there's so much information that's there. But how could such databases going forward help us to even be better at predicting disease states or potential therapies?

Yeah. I think the history of a lot of insights in medicine has actually been a product of this. It's not necessarily a new technology, but has been the product of measuring people at scale and learning from that. So whether traces all the way back to sort of the Framingham Health Study and the Nurses Health Study to understand, like, how do we know that lipids are actually bad for our heart? Well, someone at some point had to measure populations over time to actually learn that insight. And I think the UK Biobank, which you mentioned, is probably the closest that we have to measuring people at scale with modern biology methods and imaging. And we've seen all of these headlines come out of, oh, well, we've applied, you know, with AI, we've discovered how to predict Alzheimer's disease 10 years earlier using this dataset. And that often gets sort of thought of in the media and in popular discourse as an AI story, but it's actually not an AI story, it's a data story, right? Like the AI was just sort of the last final, ingredient of making sense of the data, but it was the data collection itself that enabled those discoveries to happen. So I'm pretty bullish that unlocking AI's potential across disease states, cancer included, starts with this process of how do we actually systematically collect high-quality data at scale.

So we're speaking with Dr. Amelia Jaworskyj. She's the Director of Futures at the Future of Life Institute. We're talking about the promises of AI and maybe some ways we need to refine our thinking if it's really going to shape the future of therapies. This is the Talking Biotech Podcast, and we'll be back in just a moment. And now we're back on the Talking Biotech Podcast. We're speaking with Dr. Emilia Javorski. She's the Director of Futures at the Future of Life Institute in Boston, and we're discussing the role of AI in disease prevention and maybe therapies. And Uh, maybe some better ways of revising our thinking to get in touch with reality and some of the limitations that are there and maybe ways to overcome them. And we talked previously about the UK Biobank and these other databases and how they've been useful, but it seems to me, and you can tell me your thoughts on this, that it's a great time for the USA to take the lead despite all of our current hurdles. And run this health moonshot effort. And let's just gather a massive set of data based on lifestyle, based on ethnicity, based upon, you know, what church you go to and, you know, what beverages you drink and match it to as many biomarkers, DNA sequences, and other health outcomes that people are willing to divulge. And could we potentially convince the population, which we can't convince people of anything, We can't even convince we've been to the moon, but could we potentially convince people that giving their data and monitoring their own health could come back to help them or someone they love later?

Yeah, so I'm actually optimistic about this, even though we can't tend to sort of convince people of things. And I would highlight maybe 2 reasons I'm optimistic. One is, whether it be concerning or not, people's sort of understanding of privacy and willingness to talk about health and divulge information has really radically changed since the early days of US privacy law, kind of when we did HIPAA back in the day. And we're actually seeing this with these commercial AI chatbots that are on the market. One of the number one things that people use them for is talking to them about medical questions and therapy and health-related things. which is, you know, a corporate company with no necessarily privacy protections or anything that you're just openly having this conversation with, right? So I think a structured effort that actually did have those, some privacy provisions that could be clearly communicated, I think people are ready to have that conversation for sure. But I think the second part's really important, which is needing to ensure that people benefit from participation in this. And I think this is one of the big barriers to participation in biomedical research and in general where people are very dissatisfied with the healthcare system and rightfully so. It's not working for most people in this country. And so the idea that like I'm going to participate in something and give up my information or go for extra blood draws or things, you really need to develop, close that feedback loop in a way that those results are delivered to people in ways that actually materially change their lives and improve their lives and help them access care. So I think I am totally with you that this is where we should be investing our time and effort is into a moonshot like this. And I think there are ways you can structure it with privacy guardrails plus benefit sharing that could really get this thing moving and get some momentum.

Yeah, I mean, if you really wanted to make America healthy again, if we wanted to lean on this in the correct thing, we would be developing better data that would shape our future decisions, and that as new tools would arise, we'd be building them from an evidence base rather than— well, now things seem to be done just on gut feeling more than even looking at scientific data we do have.

Yeah.

And I know I've always been really willing— 2 days ago, I got a COVID shot of an— of a— some— I'm in a trial for an experimental vaccine that should be where I guess I'm in the placebo group. I hadn't even noticed But this is the kind of stuff I like to do for duty and humanity, you know? And I get excited about the idea of contributing to data that can help people be better. But the other side of this is, I guess, the approval of new techniques and new therapies and new therapeutics. And you've written about that despite access to new tools, our system for drug testing and drug approval, it's still the same. So we can do things faster than ever in terms of the laboratory and in terms of enabled with omics and all the things we can do, but we're still dragging our heels with a current system for approval. So how could we revise that system to be more warp speed about innovation and without compromising the safety?

Yeah, I think this is one of those sort of false choices that is often put out there, which is that anything that would help accelerate or change drug approvals is somehow coming at the cost of safety. And I don't actually think that's the right way to think about it. I think the way to think about what the role of the FDA is today is to say, how do we develop a regulatory structure that can kind of keep up with modern biology and our understanding of biology? Because the rules and regulations we have now are really a product of what was happening in the '60s, right? And that was a lot of the thinking that informed our original sort of drug design. And that was this paradigm of like one disease, one target, one drug. And it was very straightforward, very rational. And that was kind of before we discovered just how complex things are and the fact that, you know, molecules working in synergy could be the right way to go or within a particular population that like it might work for some groups and it might not work for other people, right? And so you can end up with, under the current system, a drug that you see really, really worked for some people and then didn't work for another group of people. And you can find out what actually drove that effect, but that may or may not actually get approval unless the average of those 2 things met or didn't meet a particular endpoint, right? And so I do think there's a big piece here, and many people are thinking about this and sort of trying in terms of figuring out, you know, how do we help the regulatory system catch up to and continue to be dynamic and evolve as our understanding of biology evolves? And so I think that's more of a, like, updating based on understanding piece. The other sort of side of this is, like, how do you actually help things move faster without compromising safety? And there's been a lot of ways of thinking about this, which is, like, Can you actually work to get preliminary approvals to help that get to patients earlier and then have bigger requirements on confirming the benefit in the real world? And if you confirm the benefit, the drug can stay on the market. If you can't confirm the benefit, then you pull it back, right? And so that's like another modality of like how we could potentially move things, sort of move things to sort of patients quicker. I think the bigger question that sort of the shadow over all of this is the economics around this. What do payers actually pay for? And the fact that like certain disease states that are shown to be profitable, then invest more money and more money in those. And you end up with people hurting and chasing things where there's actually already good solutions, but leaving all of these areas where there's huge unmet clinical need and patients that, you know, really have no good treatments that are kind of neglected because the business model hasn't been proven yet. Right. And so I do think there's a lot of thinking that needs to happen there of like, how do we change business incentives to get scientists funding and funded and developing cures for the places where we need them? Yeah.

How does that problem affect research in things like really rare cancers especially? And that seems like that would be a real problem.

Yeah. So the FDA does have a certain pathway sort of for rare, what it calls sort of rare diseases and rare disease pathways to help accelerate approvals. I think understanding and unlocking the business model for rare diseases is actually going to be critical for all disease going forward, because to the conversation we've been having, as our understanding of biology becomes more individualized, you actually move away from these sort of large of disease to much more individualized types of, and smaller pools, right, of patients across these different categories. And so that's where I think what's happening in the rare disease community, and they're thinking about how to redesign clinical trials that can be powered for smaller sort of groups of patients and things like that, are ultimately probably going to become the norm in the future across all disease states as we realize that, like, Many diseases we think are one thing are actually probably a set of different diseases that then by breaking down those categories become something looking more rare.

Yeah. And do you imagine as a physician, what do you imagine the patient experience will be like going forward? Because I tell my students all the time that maybe as soon as 4 or 5, 10 years from now, that this is going to start with a genome sequence and a printout. I mean, how do you feel that AI will influence that relationship before, say, just a regular old doctor visit?

Yeah. So I think that that's right. And I think I would add other omics onto that. I think like your future interaction with the healthcare system involves getting your genome, getting your proteome, right? The proteome is a great tool because it's more dynamic than your genome, right? And how it works together with it. I think you can put metabolomics and lipidomics in there as well. But I think data collection will be totally different, right? Now we go in for an annual visit and you'll get a blood draw and you'll get like 10 things back, 30 things back at most, like for if you're doing one of those deep dives, you know, 50 to 100 things back when we actually can measure thousands of things, right? And that is where I think the promise of AI is really exciting is the combination of bringing our ability to measure biology, those thousands of things into the clinic, and also having sort of the computational tools around it, like AI tools to help us make sense of that, right? If you can make sense of, you know, the 30 things and your doctor can make sense of the 30 things on your annual visit, becomes a lot harder when there's so much more data involved. But I think that to me is where AI could be really exciting.

Yeah, but this is the problem I see as a bench scientist, okay? As a guy who is— the amount of data that you could potentially interrogate for as a diagnostic criterion is infinite because, okay, you got a metabolome, but where do you take it from and when? Your metabolome at noon is going to be different than your metabolome at night or after a meal or before a meal or like everything's— and in different tissues. Is it a—

Yeah.

So you just take a blood It's different than what's happening inside muscle cells. So how do we even begin to wrap our hands around the idea that coming up with these diagnostic correlations and these tools and biomarkers or whatever we're looking for, it still is going to require a tremendous investment just in the fundamental basic science, right?

Completely. And there are some great case reports out there, even in the regular lab tests we get today, actually very Quite a bit more than you would think based on if you've exercised or not, or what time of day you're taking it. And we're in this sort of circadian rhythms. And I think we've seen this too in some of the studies of these like longevity clocks, right? We're going to measure your organ age out there. And people get like, depending what day they did, they're 20 or 60, right? And so the stuff is all over the map. And so figuring out there's both the— data collection piece of it, but then there's also the data standardization piece of it. And like, how do we actually understand what variability is normal? What variability is actually a signal, right? How do we translate different ways of measuring the proteome metabolome across different clinics, sites, right? Labs even, right? So there's like a huge underappreciated data, like, cleaning standardization process that needs to happen as part of this. And this is where I think research funding comes in, right? And we look at like, what does the NIH fund? What does ARPA-H fund? What do these different domains fund? Like, we need to be supporting both like the creation of these datasets at scale, but also understanding how to collect that data, from where to collect that data, which data points are actually the most clinically predictive versus really noisy and actually totally useless. And so those are like the big questions that I think a lot of folks on the tech side of things assume is settled, right? Because there's a way of collecting something, then that's just one thing, and then we'll just do that at scale. But that's, you know, anyone who's been in the lab knows, like, you use different reagents and you get different, wildly different results. Like, the devil is really in the details on this one.

No, very true. I mean, did you let the sample sit at room temperature for 5 minutes or 1 minute? Did you put it in a refrigerator that was in the the front of the refrigerator or the back of the refrigerator where there's a 3-degree difference? All these things affect sample quality and ultimately stability of specific biomarkers that may be there. And I think the complexity of this is so much more deeper on the boring side than on the exciting side, the AI tool side. And along that same line, so whereas many folks who are interested in this are looking for cures in the data, Isn't there maybe better applications, or are there applications for AI that are much more mundane that could have substantially better patient impacts, like even things around the clinic, things like that?

Yeah, so I think this is where AI is particularly exciting for the healthcare system. And I'll just throw out a couple of ideas out there. One is on biomanufacturing and manufacturing applications. So thinking about promising new therapies like CAR-T therapy, right, for cancer. How can you actually bring down the manufacturing costs and better standardize that? And like that, it's sort of to your point, the dotting the i's and crossing the t's on the experimental side and manufacturing side and quality assurance side. I think that there's a lot of applications for AI in that domain. On the other sort of unsexy side of medicine is the fact that there's so much administrative waste in this system. And I think this is something many people don't realize is just what percent of the healthcare expenditure each year is wasted on administrative tasks or middle management or middlemen embedded in the system. And I think that's someplace where like we can get a lot of cost savings and efficiencies using AI that doesn't actually compromise care at all, or if anything could potentially help improve quality of care. And now like the huge asterisk of that is like if we choose to design it and use it that way, right? Because if you sort of thoughtlessly just integrate AI into stuff, it can kind of supercharge what you're already doing. And if what you're already doing isn't producing the results you want, then you're just gonna not produce the results you want faster, right? So those are like 2 examples I'd throw out there of like the biomanufacturing QMS side of things and also like the administrative administrative side of things where there's probably a ton of juice for the squeeze that already exists in the AI we already have. Yeah.

See, I love that topic, the idea of the CAR T-cell therapy accelerating. I was just talking to my class yesterday about this. Could we imagine a model— right now, CAR T-cell therapy or the new sickle cell therapies, they look like they work in some cases, some cases they don't. We're figuring that out. But how do we get this where you basically have a Jiffy Lube for CAR T-cell therapy so that we can decentralize this away from a regular healthcare environment and be able to do this in a way which is more patient-centric, but also cost-effective, and maybe being able to do this at scale? And I can imagine these would be places where these kind of tools could be really valuable.

Yeah, I completely agree. And I think that point you make about decentralizing the manufacturing capability is really important. And to the extent that we can de-risk and decentralize things, taking them off of a central supply chain that everyone's dependent on, to be able to do a more modular manufacturing approach that can happen closer to the patient, and whatever scale that is, whether it be the hospital, the region, what that may be. You're sort of seeing that with mRNA therapeutics as well, and figuring out how do we move from centralized to kind of modular, decentralized manufacturing. So totally agree that that is the direction where, where we should be headed when we think about biomanufacturing, is more at how do we get that to a local level and off of one large centralized supply chain someplace.

So, so you and I align on a lot of stuff here, both as, you know, scientists and, and people who are thinking about the future and kind of dreamers thinking about maybe ways the system can be better. But I guarantee you that most people in the US at least and throughout the world really want the days of Marcus Welby. They want the physician to come to your house and stick the tongue depressor in. They want a much more personalized care that's one-on-one. And we're reducing the patient to a DNA sequence and smudges on a flow strip and a bunch of biomarkers. And as a health— As AI is penetrating healthcare more and more, how can we steer it to benefit the physician-patient relationship rather than make this even more impersonal and sterile?

Yeah. So I think this is really the— if we do it right, where the— what the promise of sort of AI could look like globally, right? Not just within the United States or industrialized countries that have very sophisticated healthcare systems. But if we can really bring down those sort of measurement costs, right, and analytics costs and the ability to sort of make things locally, right, in terms of therapeutics and print drugs locally and make CAR-T cells locally, right, you start to enter a system that can be actually much more human-centric where your provider's job is to kind of be more of a companion and Sherpa of the information that is at the state of the art, you know, and like I look at the doubling rate of medical knowledge. And we're like, how we think about the role of medicine is just changing and evolving, right? Like, if you graduated med school in the 1960s, throughout your career, like, there would be incremental changes in practice, but it wouldn't be anything particularly radical. Like, what you learned in med school is still pretty much what's happening in, in, in the clinic. Whereas now, where our sort of understanding of disease and our ability to, as we said, interrogate biology and test new treatments and try and individualize approaches, that's accelerating, right? And so the role of the physician and clinician, I think, is really going to be the person that can ground that acceleration in reality, right? Like, what here is something that we should be taking forward to my patients and discussing with them? What here is just hype and noise that, like, hasn't actually been validated yet and, like, should not actually— is not clinic ready, right? Is not people ready. It needs some more work. And helping sort of people through that, make sense of all that sort of data that's coming at them to tell them what actually matters to you, what's gonna move the needle for you and all this like noise and information, what in here are the important things for you to know? And like also how can we help, right? Because I do think as much as we're gonna have fancy new ways of doing things, there are the bread and butter basics Of like lifestyle and eating healthy and exercising and sleeping, and those sort of bedrock unsexy things of medicine that you still kind of need a coach and a person to work with you on. And I think that to me is kind of the where the future of medicine and future of doctor goes is not that you know patients are just gonna log onto an app in their computer and that's gonna be it. And you know that. that's their interaction with the healthcare system is that like it is a future where you still have a doctor, the doctor helps make sense of what matters for you and what doesn't, and then is still kind of your person to be your guide and coach to get that, get you on a healthy path forward.

Now, this has been really, really helpful. I think just getting familiar with your ideas and the concepts you present have really reshaped my thinking on this in an optimistic way. I think I feel a little better about this than I did before. So, Dr. Emilia Javorski, thank you very much for your time today. And if people want to learn more about you or the Future of Life Institute, where would they look?

Thank you so much for having me, Kevin. This has been such a wonderful conversation. To sort of learn more about the Future of Life Institute, we're at thefutureoflife.org. And a piece I just kind of wrote on AI and cancer, you can find at curecancer.ai, if that's Very good.

And I'll put a link to that on the website associated with this episode so you can find it there. So this has been really eye-opening because I get really excited about technology and how technology can help us, but I'm also very skeptical in all of the promises that folks make. And I think as we go forward with a much more sophisticated and nuanced view of the limitations and the potential good things that can be done with technology, If we just have better data, I think that holds a whole lot more promise. This is the Talking Biotech Podcast.