This Week in Startups - E1089 The Next Unicorns E11: insitro CEO Daphne Koller is revolutionizing drug discovery via machine learning & data, shares insights on remote education from time at Coursera & more

Episode Date: July 29, 2020

Follow Daphne: https://twitter.com/DaphneKoller Check out insitro: https://insitro.com/ Follow Jason: https://linktr.ee/calacanis

This week in startups, the next unicorn series, is brought to you by LinkedIn Jobs. A business is only as strong as its people, and every hire matters. Get $50 off your first job post at LinkedIn.com slash unicorn. Zendesk. Qualifying startups can join the Zendesk for Startups Program and get six three months of Zendesk products. You'll also get access to an exclusive. community of startups for advice and connections, visit zendesk.com slash twist today to get started. And Vanta.

Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. Hey everybody, welcome to this week in startups. It's our next unicorn series where we look for companies we think that could change the world and get really big in doing so. And it's really interesting to have our guest back on the program when I first met her about, I guess, seven years ago. She was doing a company called Coursera, which of course you know is doing massively online courses. MOOCs, I think we called them back in the day.

And paradoxically, you know, we now have. everybody doing Zoom and working from home, so we'll get into that a bit. But Daphne Kolar is an expert on machine learning, computational biology, I believe, was an AI computer science professor at Stanford, and she's got a company she's been working on called Encitro, and they are solving perhaps one of the the even bigger problem than education, which is in biology and in drug discovery, unlike doing artificial or machine learning on text, which is really easy. You got the whole internet. You got the Wikipedia. You got a million different sources, books. We don't have big sources of data. We don't have big data on biology. And so her company, EnCitro, is trying to solve that problem. Welcome

back to the pod. Seven years later, Tiffany Cole. How are you? Good. How are you? Very glad to be back. Yeah. So before we get into what you're doing at in Citro, the irony must not be lost on you that seven years ago you were part of the groups that were saying, hey, all of this knowledge should be online so that people could take remote courses. And of course, people laughed at you and said, oh, well, that doesn't work. Online courses, that's never going to work. And here we are in the summer of 2020 recording this. And we're all watching the top universities try to figure out if they're going to be coming back to school. So, What did you learn about Coursera in terms of remote education?

And what do you think about these, you know, in some cases, Ivy League schools, having students pay $50,000 to take an online course or a series of them? Or almost entirely online courses. No, I think we were definitely ahead of our time in trying to launch an initiative like Coursera. I think we were successful in many of the audiences that we tried to hit. So a lot of the learners who were on Coursera were people like you and me who'd finished their official education and were in the workforce and really felt the need to learn something new that would help them advance their career or just expand their mind.

And I think those were people that were well served by Coursera back in the day. But I have to say, we didn't make a big impact on how universities could teach their traditional students differently. And I think everyone was kind of feeling like, well, it ain't broken so we don't need to fix it. And while I think it's true that it wasn't entirely broken, there were clearly ways in which the teaching and learning process, especially in the large lecture classes, could have been improved at the time. But people just weren't willing to pay attention. It was too much work and students were doing okay, so why bother? This pandemic has forced us to really evaluate this.

from a new perspective, and we're at this point, at the point that there's just no choice. And what I'm hopeful will happen out of this is because people are now forced, universities are now forced to teach differently, maybe they'll realize that there is actually a better way. And that will make that long overdue transition finally happen. Yeah, and we see that in work. We had a whole group of people, these weird people running remote companies, and we kind of were looking at them going, that'll never work. You can't run the next Google or Facebook, Coursera, working from home by Lake Tahoe,

wherever you are on some beach. And now we're all doing it, and it's working. And many of us are lucky to work behind keyboards. So it actually, in some cases, if you learn the techniques work better, looking back on Coursera, was a beneficiary of the effort to put all this coursework online, people who, like you're saying, were maybe left out of these courses, ultimately, as opposed to the people who are on the university campuses, those were the beneficiaries?

I think initially that was definitely the case. These were people all over the world, in fact, who would never have access to an education, at Stanford or Princeton or Michigan, and they were the ones who really all of a sudden had these incredible learning opportunities that were not previously available to them. The audiences that we really didn't hit,

were the ones that had that opportunity, but I think in a way that was not as good as it could have been. And hopefully we can now give them an even better education by having been knocked off of our little hill and hopefully be able to climb a better one. What did you learn in terms of what makes a good online learner as in terms of the students, a teacher, and in terms of the platform?

What gets people to actually learn when they're on their computer alone in their living room or their little study space? You know, honestly, that's very hard. And the ones who learned the best were the ones who had some kind of extrinsic motivation. These were oftentimes ones who had some clear goal of what they were looking to get from the online courses, like they were looking to get some kind of certificate or such that would allow them access to a better career. The other impetus that we found was really effective was some kind of online community so that people felt like there was almost a series of deadlines and they were beholden to other folks who were in the community with them to make progress and complete the course.

And if we're looking for what I think would work well in the university setting, it's not to put a random person in front of a computer and say, hey, here is the entire course completed at your own pace. let us know when you're done, because I think it's a relatively small group of individuals that can actually achieve that level of self-motivation. What you really need is to create some kind of community with a synchronous experience that keeps people moving along at a certain pace. So I think if I were a university now, it would be looking to have some blend of really high-quality online content, including what we find on Coursera, along with a teacher

and a community of students who are kind of working through that content together and making sure that there is a discussion, there is engagement, there's sort of a forward pull towards completion because otherwise people just tend to drop off and then they never come back. We are pack animals and like when there's a cohort and it's social,

it makes you come back, right? And the gamification also, it's really also interesting when you think about it. we've seen another layer of more casual education, I think, that learned lessons from Coursera and that cohort of MOOCs, which are things like masterclass or, you know, other online learning opportunities. But you can confirm that you did not solve human motivation to learn. No, but I think that's a place where the expectations were just misaligned with reality. So when I pick up a book by Thomas Friedman or I see the lean startup behind you and I pick that up and I read three or four or five of the chapters and I feel like I've gotten a lot out of it and I don't finish the book.

Is that a failure of the book? I don't think so. I think the book provided a lot of value to me and that's terrific. So I think there was a misalignment in that people expected that. this to be like a college class in which completion corresponded to success. And to me, for those casual learners, the ones who are just looking to expand their mind, success is walking out feeling like you've learned something of value to you. And I think that was part of the narrative around MOOCs that we always tried to correct and say, look, this is not like a college class.

And you shouldn't expect people, most of them didn't even want to complete going in. They just wanted to learn something. And that's really the question that you should be asking. Right. Fascinating. That people want to come in and dabble a little bit, get a little bit of knowledge, and then maybe come back a little bit later, you know, like people who bounce around from books. It's a thing that's a really good analogy. You think we can safely go back to college and then what is the future of the university post-pandemic? Do you think we return to the regular world in 2021, 22 college season? Or is this a permanent reset in some way? Even if the, let's assume the pandemic is resolved in the first half of next year, 2021, do you think college has changed and people

look at them differently? I think the answer is yes, it's going to change, but it's not going to go away. So I have a daughter who's about to start her freshman year of college. And I can tell you that she is bitterly disappointed that at least for now her college is going to start online because it's not going to be safe for her to go back to college. And I think for a lot of people, the going to college is way more than just what it is that you learn. It's about going and being part of a community, forming a new social network maturing as individuals. And I don't know that you're going to get that just by leaving home and living in an apartment somewhere on your own, which I guess would be the other option that she's considering now. So I would say there will be a college experience, but I'm hope,

at least for the undergraduate population, but I'm hopeful that that college experience will be better than it was before. So for instance, I never thought, and for me this was actually the entry point into my whole Coursera adventure, that the in-class experience of a large lecture with 500 people sitting some of them way back in the back row and having this little sage on a stage at the front lecturing at them for an hour and 15 minutes, that's a really great learning experience. It's just not.

And all of the studies prove that that's not how one learns effectively. But we've kind of converged on that as the sort of, this is how we do things, even though we all knew that it wasn't quite right. So I'm hopeful that because we're forced to get a new generation of tools that will allow for more of an interactive study group experience and, you know, students talking to other students and working on things together. Because, you know, replicating the online lecture, sorry, the big lecture hall experience in a video format is absolutely horrendous. I mean, if the in-class experience there was not great doing it on video is much less engaging.

So we're going to have to develop a new way of doing things. things and that might be better than the large lecture hall. Now, honestly, I'm not sure that for some other educational experiences, it will end up going back to normal. So if you're a master's student and you're already in a professional environment, you're already working, is it really necessary for you to pick up, uproot yourself, your family and go off to school or is it better to just learn online. That's where I think we might not see quite as much of a return to face-to-face teaching. The continuing education, the graduate degrees, because you are obviously at that level very motivated and you're an adult and you can drive yourself to do the completion, right? I mean,

it's a different group of people. And you already have in many cases the things that undergrads are looking for. You have your social network. You have an apartment. You have a job. You don't actually necessarily want to pick yourself up and go live in a different location just so to you can attend to the university there. All right. When we get back from this quick break, I want to talk to you about the new company in CITRA, which you've been working on for a couple years now, two or three. I think you started in 2018.

And why you started it and what you hope to accomplish when we get back on the suite of startups. Listen, we all know that LinkedIn jobs is amazing, but I wanted to start today with a testimonial from one of you, the audience, who recently emailed us and told us, as of this week in Startups listener, about their amazing experience using LinkedIn jobs. Well, Aaron Mason is the founder and CEO of Emma AI, a startup that uses AI to optimize travel time on your work schedule. And Aaron recently hired a machine learning engineer who is starting on Monday. He received 110 relevant applications in only four days.

That's over 25 a day. And he did that with a small budget, so he got exceptional value. From job posts to offer accepted in only a few days, LinkedIn jobs for the win, of course. and you know small businesses have unique needs, despite the current uncertainty. One thing that remains unchanged is the importance of having the right people on your team. You know this. So when you're ready to make that next hire, LinkedIn jobs can help by matching your role with qualified candidates so that you can find the right person quickly.

That's what you want. Quality and speed. And LinkedIn does that. They have an active community of professionals with more than 690 million members worldwide. It is ginormous. and they put your job post in front of qualified members every day so that it's seen by people looking for jobs like yours. And they might even put it in front of some passive job seekers.

That's how LinkedIn jobs helps you find the right person faster. So I want you to get that $50 right now at LinkedIn.com slash unicorn. That's right. Go to LinkedIn.com slash unicorn and you will get $50 off your first job post. Of course, terms and conditions apply. Thank you, LinkedIn, for giving us $50 for our listeners. at LinkedIn.com slash unicorn. All right, Defny Kolar is back with us on this weekend startup to talk about her new company

in CITRO. So tell us, why did you start this company? And what do you hope to accomplish with it? So I've been working in the area of machine learning applied to biomedical data sets for quite a while now. I think it was in the late 90s, early 2000s. And I always felt like there was so much. richness, so much potential value in biological data, because biology is one of the most

complicated things that we have to deal with. It's just, it's not a system that we get to design. And so the rules are really complicated and intricate and often evolved. I mean, in general, evolved to where they are by a set of random events that created something that is quite magical, but also incredibly complex. And the only way that we, I think, are really going to understand the biology of who we are and not just us, but nature around us, is by collecting enough data and applying sufficiently sophisticated algorithms that can really extract some meaningful rules and insights about what's going on.

So this is something that's fascinated me for about 20 years now. But it was always really hard to do that because data collection in biology involved weeks, months of wet lab experimentation by people sitting there and pipetting and moving liquids from one tube to the other. And so if you got a handful of data points, you felt lucky. And now that world has changed. And we are in a position where there has been a set of tools that. have been developed by bioengineers and cell biologists that all of a sudden give us the opportunity to measure biology at unprecedented fidelity and scale. And that is just an incredible opportunity for people to bring that suite of tools together with tools from machine learning and data

science to solve problems that are at the core of society, at the core of making the world a better place. I think there is opportunities like that across multiple domains. I think it could transform agriculture and crop growing. It can transform the environment in allowing us to create organisms that clean up our oceans, for instance. And importantly, from what I want to do, it allows us to understand and transform human health by making better. are drugs that help us deal with diseases that are currently intractable and cause tremendous amounts of pain and suffering and death because we haven't had the tools to really probe into the biology and figure out how to fix them. And that's really what we hope to do it in CITRO.

And so these, the machine learning, if I'm correct in describing, and I always try to describe it in the simplest words possible is software that learns from the data that's presented to it. Is that a good simple definition? Fair enough. Yeah. So the data sets don't exist. Does that mean you're making the tools of machine learning or that you're preparing data sets and the tools so that a drug discovery company, you could sell them the tools and say, hey, here's a bunch of data. Why don't you look into it and see what you can find? So you have a reference set. almost except up until you got to the last bit so we are making data we are also building tools both tools to create data and tools to interpret data but we don't sell tools we are a drug

discovery and development company we just want to be the new generation of drug discovery and development companies so in the same way that when you know companies went into the retail space and they said, we're going to build, thinking of Amazon, for instance, a completely new way of doing retail. They didn't just take Sears and said, let's transform Sears by putting a website on front of it. They said, we're going to build an entirely new way of doing retail. We want to build an entirely new way of thinking about drug discovery and development where data science, machine learning and large data creation just permeates the entire process. So you're building the tools that you need to do the drug discovery that at a much more

competitive or cost-effective way. Is that what it's about getting more swings at bat? Because my understanding is drug discovery is getting more expensive. And it's getting harder. That doesn't make sense to me with Moore's law, with software, with Amazon Web Services, cloud computing. I thought everything got cheaper. Why is drug discovery more expensive today?

So people will give you different reasons. Partly it's that there's going to be some people who blame regulatory in the FDA for making the recording and clinical trials more onerous. And I think that plays a small part in it. But I don't think it's the fundamental problem. The fundamental problem is that most drugs fail. And success rate for a drug that, for drug program from beginning to end, depending on exactly how you calculate it and at what stage you count the beginning, is about 95% failure rate. That's not 95% success rate. It's the other way around. Sort of like startups, yeah. Yeah, only, yes, exactly. And so that's a serious issue, because each of those drug programs is incredibly costly.

And you only discover often at the very end after you have invested a tremendous amount of money and time that the program's not going to succeed. And so each of the successful programs carries on its back the cost of the, you know, 20X programs that didn't succeed. So what we hope to do is to build predictive models that tell you much earlier in the process that certain paths are going to be dead ends. And so you can avoid going down those paths and incurring all those costs that add on top of the cost of the successful programs. So this means you get more swings at bat. You could even be more risk-taking. And, you know, even in the 20x example you're giving, that actually, you have to return more than 20x because

capital allocators have a choice of where to put money, as you know, as somebody who's been in the startup world now and taught at Stanford with is a ton of entrepreneurs. Not only as you have to make up for those 20, you also have to then beat the stock market's return of 7, 8, 9% a year. if a capital allocator like a pension fund or a venture capitalist is going to fund these things. So who funds these drug discoveries now? And what is the scale of them? Is it $5 million to $50 million or $500 million? Educate me on the scale of discovering a drug and then getting it commercialized?

You know, it really depends tremendously on which disease area you're going after. So if you're going after an area where the trials need to be large because there's many aspects to what makes this expensive. So some trials need to be really large because the population of people who would potentially be taking the drugs is very large. Cardiovascular disease is one example. Vaccines is a really good one. And so you need to make sure that you've done your trial. large enough so that if there is a rare negative adverse effect, like a really significant risk, you discover that in the clinical trials before you put it into, in the case of a SARS-CoV-2

vaccine, into 7 billion people, because if you don't discover that side effect that's a one in 1,000 or 1 in 10,000, and all of a sudden you multiply that 1 in 10,000 by 7 billion, that's a lot of people who can be at serious risk. So that's one aspect of it. The other aspect, of what makes some of the programs more expensive is that you need to make sure that you have enough people to be powered to see a significant effect. And that's an area where I think a lot of drug companies make a mistake by trying to go after very large audiences because they think it increases their market. But that basically dilutes the population where you're really going to make a big difference.

there's been a movement in the last few years towards smaller drugs that target a smaller, much better defined population where you can really make a very big difference from the perspective of almost a cure. And they're not blockbuster drugs as they used to be when we did statins and such, but for the patients for whom they work, they work incredibly well. And that's where I think the field really needs to go is drugs that work for a smaller population and really, really make a difference. However, with a small hit rate, that's impossible because who would finance a mission of bringing back a small prize, even for a small population?

So how much of this is the people financing it just saying, you know what, I need you to go swing for the fences? And the entrepreneurs who are working in these companies saying, we need to swing for the fences? I think there is actually a big bifurcation. here between smaller companies and larger companies. So a lot of the smaller biotechs have actually been quite successful by not swinging for the fences in the sense of going after the large populations, because for them, if they get a 3x, 5x return on an investment and that, I mean, it can still move the needle dramatically in terms of the benefit that they provide to

themselves and their investors. Oftentimes it's actually the big pharma's where, a smaller hit doesn't move the needle. Got it. If you're a, you know, if you're a 50 billion, 100 billion company. Right. And so for those, it's oftentimes a pressure to go after the bigger indications rather than slice it up and really make a difference to smaller groups of patients, but a very

meaningful difference. And why aren't those big companies using machine learning themselves? what why does your firm need to even exist? Are they just, they don't have that skill set as their core skill set. And this is just two different topics, you know, machine learning and,

you know, drug discovery and building drugs. I think that's a really great question and one that I get asked a lot because everyone knows that they need to be doing machine learning. All of the CEOs of the big pharma, they go to Davos and everyone's talking about how machine learning is changing all of the other industry.

and so they come back and tell their people, hey, we need some machine learning. Can you get me some of that? Yeah, sprinkling it on. Exactly. And the truth is that I think it's really hard to take any company, not just a pharma company that doesn't have a data culture and suddenly switch it to be working on a completely different model. So, for instance, in a lot of the pharma companies, there's this incredible siloing of the data that they've collected. Data was never considered an asset.

It was a disposable. So it was stored on random disk drives and different people's laptops. I'm not kidding. I mean, it's not like there's in a ETO. No, no, most of them are still on-prem. Most of the, a lot of the data are still stored on laptops. And so in order to even amalgamate that and curate it and organize it in a way that makes it accessible is really hard. And it's years and years of work.

that often people are not willing to put in. And when they do, the data is often obsolete. The other aspect is the cultural aspect. So think of the shortage of machine learning engineers out there. I mean, we know that they're worth their weight in gold right now. And if you're a machine learning engineer and you go into like a big pharma company, you know, they don't oftentimes value those folks in the same way. that, you know, if you go work for a big tech company.

So they don't have a seat at the table in terms of making the decisions about big strategic initiatives. They're kind of summoned in when, okay, we've done the experiment, generated the data. Could you please go analyze it for me? And that's not a very rewarding role for a talented engineer. So a lot of them are just going to go work for a tech company instead. Yeah, we're there the stars of the show. It's sort of like you could, if you were a fashion designer, like would you want to be the

costume designer in a movie. Like, you're not as important as the director of the stars, the person who wrote the screenplay. You're just kind of like, ah, just bringing the set to bring in the costume designer, change the costumes. But you want to be the star of the show. And those engineers are going to go work at Google or, you know, Tesla and work on self-driving cars or something that is super rewarding and go work for your company, which would also be super rewarding when we get back from this break. I want to know where you're aiming, you know, this machine learning gun and what you're trying to take out. What are the targets when we get back? back with Defny Kolo of Encidra.

All right, things are a little weird these days. We all know that, but now more than ever, it's important to build and maintain great customer relationships. You know that. That's the heart of every startup. You need to have a great relationship with your customers. And Zendesk is here to help with their Zendesk for startups program. Yes, that's right.

Zendesk is the first service for a CRM company with support and sales and customer engagement products designed to improve your relationship with me. customers and the Zendes for Startups program is offering qualified companies six months free and you'll be able to utilize their support and sales solutions and gain access to an exclusive startup community with resources to help you scale out your customer support and here is one of my breakout companies that loves Zendesk you know Steasy studio they went through our accelerator they teach you to dance my wife's addicted to it everybody loves stezi I get really great comments on the product all the time well through a combination of Zendesk Explorer

and their ticketing tagging system, they can track which features their users are most excited about. And Steezy can then relay that information to the product team. So for Steezy, Zendesk is more than a CRM tool. It creates a positive relationship with members and empowers them to contribute to Steezy's growth in return for some awesome dance moves, which I will not break out at this time. I will save you for my dad dance moves. If you're an early stage startup, here's your call to action. Early stage means you have under 50. employees, well, you can get started today with six months free. All you have to do is have under 50 employees, which I think is most of you're listening. Go to Zendesk.com slash twist.

Every customer counts when you're a startup, especially now. So start building the best customer experiences with Zendesk. And they've also launched their new podcast. It's called Sit Down Startup, where Zendesk leaders chat with founders, CEOs, and makers on their startup journey. You can listen and subscribe to Sit Down Startup on Spotify or Apple Podcasts right now. Just search for sit down startup. Thanks for supporting the programs, Endest. All right,

welcome back to this week in startups where I'm getting a quick education on, along with all of you, on drug discovery. I mean, many of the people

in the audience that you all know about machine learning. It makes total sense to me just the same way Elon is reusing rockets. That lowers the cost to going to space.

The more the cheaper it is to go to space, the more missions you can do, the more satellites you can put up for his starlink, the greater the chances of going to the back to the moon,

to Mars. Makes total sense. You're going to make it cheaper, faster, better. You're going to avoid mistakes by using machine learning. You need all those talented people who are the rock stars in your company. You've been at this for two or three years. You've obviously built some software. Are you at the stage now where you're targeting specific drugs or targeting specific diseases? What are you thinking about in terms of where to point, you know, the super weapon you're building? So that's a great question, and we did spend most of the companies two years old. We actually had our birthday a week ago. So, yeah, we're very excited about that.

And we spent a large part of the first two years building the foundation because if you're building a data generation engine, that requires a lot of infrastructure. And biology is messy. It's complicated. Cells are actually living beings and they don't do what you tell them to do. And so there's this incredible amount of work to making all of the protocols and all of the materials that you're working with as locked down as possible so that the experiments work the way you hope they would work. So now there's still a lot of foundation building that we need to do, but we've done enough that we can now actually start prosecuting diseases in a meaningful way. And we've been working for about a year on a project that is joined with Gilead,

who was a great partner to us to get things off the ground because they gave us a bunch of money that helped pay for the platform development and also some great expertise in the liver area. And so with them, we're working on a disease called Nash, which stands for non-alcoholics, teato hepatitis, which is basically an inflammation and fibrosis of, the liver, scarring on the liver, that is becoming much more predominant these days because of obesity and insulin resistance. And it's going to become the largest cause for liver transplants and liver cancers in the coming decade. So it's a real serious issue. And so we've worked with them to allow us to create cell types from people like normal,

people like you or me or people who are sick with Nash in order to understand at the cellular level what the genetics that give rise to a higher proclivity to getting Nash versus a lower proclivity to getting Nash. And then how do you drive those cells using environmental factors more or less towards a Nash outcome? and then what drugs might be the ones that revert that disease phenotype back towards a normal state. So we've made tremendous progress on Nash, and now we can, with that infrastructure in place, because once you know how to work with liver cells, then you could apply this to a whole range of other liver diseases.

The other area that we've made a significant investment in is diseases of the central nervous system. And there, of course, I don't need to tell you about the incredible unmet need that exists in both neurodegeneration and neuropsychiatric disorders, both of which are, have been considered rather intractable areas to work in with tremendous numbers of failures and the few, very few drugs that have been approved, mostly being symptomatic at best and often with significant side effects. This is an area where I think there is tremendous potential because there has been a much greater understanding of the genetics of those diseases and realizing that, you know, someone, for instance, has a neuropsychiatric disease, a large portion of that is probably genetic. So, for instance, the concordance between identical twins in autism is that. about 60 to 70% and depression is about 50. And so there's clearly a strong genetic component there that we now, I mean, it's very complex. We know that there is hundreds, if not thousands of genes that are involved in these disorders. But the fact that it is genetic means that we can see some of that in cells that are derived

from people with or without that phenotype. and you can actually look at a cell and distinguish whether it comes, for instance, from a patient with a certain type of autism versus one that isn't. And that's amazing. Yeah. That is just phenomenal that you could actually identify that and then build a drug to keep that from happening. Yeah. Wouldn't that be amazing if you could actually do that? I mean, and you can see that from these cells that are derived from basically stem cells.

from those patients and they grow, they mature in the test tube. And all of a sudden you can see, for instance, that there is a deficit in pruning of the synapses in those neurons. And that actually coincides both with what we see in post-mortem brain samples. And it also coincides with our understanding of the disease that autism is oftentimes a disease of sensory overload. And that happens because there's too many synapses that connect too many neurons to other neurons.

So what if we could reverse that process and bring synaptic pruning back to normal levels? Wow. And that could happen from a pill or from an injection. Who knows what the delivery mechanism would be. I mean, I don't want to say that we've solved this, but that, I mean, once you can identify a phenotype that you can potentially screen drugs on, that gives you the opportunity to try and find a drug because otherwise you resort to what, experimenting in people, which of course you can't do or experimenting in animal models of the disease

and mice don't get autism. They do not get Alzheimer's. They do not get Parkinson's. So you're basically experimenting in model systems that bear no true relevance to the underlying human disease. I have a kind of a silly question and forgive me if it's, yeah, if I seem really stupid for asking it.

I'm sure you're not. People do very dangerous things in the world on a regular basis. They opt into doing them, whether it's riding a motorcycle or being a deep sea scuba diver. Right? And in some cases, they do these things because they get compensation for doing it. For example, somebody who is a deep sea welder. I think that's probably one of the most dangerous jobs in the world. There's tons of dangerous jobs.

People pick to be an astronaut or maybe they fly experimental plans for a living. These are high-risk activities. And then I hear about people in science were very concerned, and rightfully so, about the adverse effects of testing on humans, on an individual human who is making an individual decision to opt into it. So I know somebody who actually opted into one of the vaccine trials for COVID. And they're getting paid $2,000, and there's 30,000 people in the trial. And I was just thinking about it in relation to COVID, there is a dollar amount in which most people would take some risk.

But it's very uncomfortable to talk about this. Right. And so what is the hang up in science about talking about people taking risk in order to try to save the rest of humanity and do something brave? We send people off to war to fight wars and they die. And we are going to send some people off to take this vaccine and some number of them could have adverse effects. they could perhaps, you know, have damages and unintended consequences. How does the science community get held to such a high standard? Is it too high?

And do other countries look at it differently? So I think it's actually pretty universal, at least in most countries, that you want to not subject people participating in these trials to unnecessary risk. And I think this has been particularly controversial in the case of, COVID-19, because there has been a pretty substantial group of people who have volunteered for what are called challenge trials, where those are ones in which the people who got the vaccine are deliberately exposed to COVID-19. And obviously, there is a non-trivial risk of that, even though you would pick presumably people who were young, healthy, had no pre-existing comorbidities and so on and so forth. But there's been a lot of discussion around the ethics of that and the arguments that the people who volunteered have made are exactly the ones that you just talked about, which is you let people volunteer to go to war, or to go into space. Why not let me volunteer to risk my life for others by doing this vaccine trial?

honestly, if I were asked, if it were my decision, I would say if someone is who is mentally competent, who is willing to take that risk, understanding all of the consequences, then I would say that's a very worthwhile societal thing to do and we should let people do this, having had the risks explained to them. But I think doctors are raised on the you shall do no harm ethos, which I think is really important. And it makes it hard for them to accept that kind of volunteer approach. And, you know, I think different people have very different attitudes about this. I mean, I'd look at it and I say, well, there is a financial equation here for a human that people make all the time in terms of what's a dangerous job. And then there's the benefit to society. So there's, you know, all these risk rewards that we do every day. And it just seems like even broaching this discussion, somebody could clip this and say, like, I'm saying people should sell their organs or something.

I'm not saying that. But, you know, it does make sense that if, and you saw this with these challenge trials, that these people are very brave for doing it, they might very much want to be a hero and do these kind of things. And if they were compensated at some very large level, boy, that could be an amazing, a deal for humanity, right? Like, we're, because we don't want to have one person die from some of these trials, these challenge trials, we would rather see millions of people suffer from the disease every year. I mean, that's kind of what you're saying.

It's, it's very hard, I guess, in the, to look at this holistically. I think people find it very challenging to, to really think through situations where one deliberately puts another human life at risk. And even though implicitly, we do it every day. And I used to teach decision theory when I was still a professor at Stanford and point out that whether you explicitly acknowledge it or not, you make, I mean, society makes decisions all the time that put people knowingly at risk without, but just don't think through the consequences.

So, for instance, when you make a decision that you're not going to. to do a full-blown maintenance on every airplane after every flight. There is no question that that would increase the safety of airplanes. It would also increase the safety of your car if you were to change the tires every four weeks rather than once every four years. And yet people don't do it, even though it is clear that by doing so, you would improve the safety of that mode of transportation. But somehow people refuse to do the calculus, even though implicitly that trade-off of money or time for risk happens implicitly all the time. It's so weird, and it's, I think it's particularly in this culture right now where people

are trying to, the over 10 window has been shut so tight, right? Like what we're allowed to talk about, this has got to be one that gets you in hot water in a scientific community, or do people actually have these discussions in the scientific community around like COVID-19 and those tests? They definitely have had, there's been ongoing conversations about, the COVID-19 challenge trials. And in fact, I believe that one of the vaccine candidates, one in the UK, is undergoing or will soon start a challenge trial in parallel to a traditional clinical trial. Now, I think we need to recognize that for vaccines, especially, the challenge trial only gets you so far.

It gets you, it gets you to efficacy. It does not get you to safety. And that calculation that I mentioned before, that even a. factor of 1 in 10,000. But then again, at that point, you also have to think about what are the tradeoffs between sending a vaccine out there that's going to cause harm to 1,000 people versus the people who would die if the vaccine, assuming that it was efficacious were not provided to them, which is why I think it's smart to do the challenge trial, because if a challenge trial demonstrates incredible efficacy,

I think we would have a serious conversation of whether what level of phase three trial that would speak to safety, you would need to do before you were willing to give that vaccine to the population at large, recognizing that it does increase the risk of adverse events, but that COVID-19 also increases the risk of adverse events. let alone secondary conditions that occur, you know, the economic turmoil, and then people being

malnourished or not having jobs, suicide, opioid abuse, and overdoses. And it's so weird in our society that we can't have this like full ranging discussion. We get back from this final break. I want to obviously, you know, the major killers right now, putting suicide aside, which is just mind blowing to think that suicide has become a major cause of death. in some ways because we're living longer. But Alzheimer's, diabetes, cancer, cardiovascular disease, I think these are the big four killers. I want to talk about in our lifetime, and you and I are of a similar age, I think Gen Xers, what would we see first, you know, in terms of solving for those diseases in our lifetime?

And what do you think about learning what you've learned if this trajectory continues, what would life expectancy look like for our children and their our grandchildren when we get back on this weekend startups? Okay, if you don't have your SOC2 compliance buttoned up, you can't close major enterprise customers. It's really that simple. And if you already have your SOC2 report, don't you want to make it easier to maintain it year after year?

Well, Vantus Compliance Software makes it easier to get and renew your SOC2. Their software continuously tests against technical and non-technical SOC2 requirements, and they partner with over two dozen audit firms who have been trained to file SOC2 reports directly within Fanta. The average 20 to 50 person company is SOC2 ready with Vanta in just two to four weeks, compared to three to five months without Vanta. And with Vanta, you can connect your tools and infrastructure and continuously monitor for risk and vulnerabilities. You can also fix issues fast with actionable guidance. You want to avoid all of those anxious on-site visits from auditors

and never again have to prove compliance with manually captured screenshots. And you're going to do that by using Vanta, V-A-N-T-A. Companies like Notion, Latest user testing, and hundreds of others have successfully gotten their SOC2 reports done with Vanta in weeks, not months, including my portfolio company lead IQ. And they told us that they couldn't imagine having to go through SOC2 without Vanta. Here's your call to action. Unlock those sales and give your employees all of their time back in their calendar to work on more important business critical assignments.

You know you want to do that. So they're giving you, and I know a number of you have already signed up. So thank you for using the code because I want to make sure that you get a $1,000. That's 10-100s. That's 20-50s, $1,000 discount on their subscription at Vanta.com. slash twist. That's Vanta, V-A-N-T-A-com slash twist, for I kid you not, $1,000 off. Thanks to Vanta for supporting independent media like this weekend startups. Let's get back to this amazing program. All right. If you're hearing my voice, you will die. Not today, hopefully, but at some point in

your life. And it's one of the things that we think about a lot, especially when you become a parent, or maybe you hit a certain age. I'll be hitting 50 this fall, which is mind-blowing to me. and just thinking about death and life extension, if we were sitting here 200 years ago, and we said, wow, you know, people are living to 70 and 80 years old. And the reason people are dying in many cases is because they've opted into it

by overeating, smoking, alcohol, drugs, whatever, and mental health were becoming some of the leading causes. I'm curious what you think about, those four main causes of death of the top ones, which one of these do you think is going to be the easiest for us to solve? Obviously, Alzheimer's in the brain seems incredibly complex. And I always hear from people that that one's just too hard. But what about the other ones?

And where do you rank them in terms of how difficult they'll be to solve for? So I think part of the confusion is that each of those things that you mentioned is not actually one thing. It is multiple things. and cancer is not going to be, you're going to find a magic bullet that has somehow eluded us. And when we find it, cancer is going to be gone. I wish it were that simple, but it's not. So I think what we're going to see is a process by which we just continually chip away at segments of these diseases and hopefully make a big dent in some of those segments and over time more and more.

And even in cancer, we've seen a tremendous amount of progress over the last, over the last decade. Diseases, some types of cancer that used to be a death sentence like melanoma. I mean, melanoma used to be like, you know, you got melanoma. And, I mean, if it was metastatic and you were dead, there was no cure. And now with some of the checkpoint inhibitors, the immune checkpoint inhibitors, there's now, for a lot of people, melanoma, they just, They're cured, which is not a word that anyone ever thought to apply to metastatic disease, right? It's not everyone. I mean, the response rates are still, are still, you know, 20, 30 percent in many cases.

But there are people who have, I don't know if they're cured because we never know if there is cancer cells that are still hiding there somewhere. It might pop up in 20 years because we haven't followed patients for long enough. But for all intents and purposes, years later, they don't seem. to show any signs of the disease recurrence. And so I think that's amazing that we've been able to do that. We haven't done that for all cancers. I mean, there's some that are still incredibly a challenge, and we haven't found anything along those lines,

like glioblastoma is a good example. I think the same thing will be true for Alzheimer's. I'm confident that Alzheimer's is not one disease. There's multiple pathways that are involved, And I think part of the problem has been that we've tried to treat it as a single thing. So my hope is that as we have better tools for studying the brain, we will identify those segments and be able to bring to bear a much more targeted arsenal for those segments in the same way that we've been able to bring to bear targeted therapeutics, what's called precision oncology,

to certain subtypes of cancer. So it's not that we're going to beat one and not the other. We're just going to chip at all of them. Right. And there's different tools, right? Because we have CRISPR in terms of flipping, you know, switches in our DNA, correct? And then we have stem cells where people are able to, is this correct that we're starting to be able to manipulate stem cells in human beings to then regenerate, at least in test tubes?

and stuff like that? And those solutions? Yeah, I mean, some of the interesting, very early studies suggest that you can potentially give, for instance, Parkinson's disease patients whose dopaminergic neurons have largely died off. You can basically transplant stem cell-derived dopaminergic neurons, and they start producing dopamine, and they give these people a better chance. a much longer life than some of the treatments that we've had, which have basically largely been

dopamine replacement using al-Dopa. So there's some really exciting treatments on the horizon, but I also wouldn't dismiss plain old, whatever, small molecules and large molecule therapeutics, because some of the biggest advances that have come in treating disease have come from taking those traditional tools and just applying them in a much more intelligent way. So one of my favorite, examples of that is the remarkable progress that's been made in cystic fibrosis, where there is a single gene. It's called CFTR. And by understanding the genetics of that one gene and saying, this is the mutation that causes cystic fibrosis in 30% of patients, we can come up with a small molecule drug. Small molecules are like the oldest, most traditional form of drug. But by designing it in a

very careful way, all of a sudden we've basically caused that protein to fold correctly and do the right thing. And now 90% of cystic fibrosis patients have an effectively normal life, whereas it used to be up until not that long ago a death sentence at the age of 20. So it's not... Interesting. You mentioned that I have a family member who's the same age as me, or approximately so, who is a beneficiary and who has cystic fibrosis and has been a beneficiary of these drug discoveries. So it hits close to home. What about these organoids and this ability to make many versions of, you know, from stem cells of our organs? You see that this eventually, because we could test on those. Explain what organoids are, why that's important. And if at some point,

you were talking about, you know, brain cells, being able to build brains and people, tree dishes. I know we can build little tiny ones. Yeah. So I think we need to distinguish between the use of those tools for research purposes and the use of those tools for therapeutic purposes and they're quite different. So the use of these tools for research is something that I think has come a tremendously long way, including the work that we do it in CITRO, where the idea is that I can take a skin cell from you with your genetics, revert it back to stem cell status. So now it's a stem cell from your genetics and we can now cause it to become a neuron. How does that happen?

Is there some liquid you pour on it? Is it some process? There's different techniques. Some of them are small molecules that kind of push the cell this way and that way and this way and that way. Sometimes you actually genetically intervene in the cell by having genes that turn on during development and really causing them to turn on at that point in the, development of the organoid. And so now we have an incredible collection of organoid protocols that

have been developed. Some of them are for brains. Some of them are liver, gut, heart, kidney. It's actually quite remarkable to see them. And you can kind of see the little brains in the dish. They're tea little brains in the dish. They have electrical activity. It's in the Wikipedia page. And it's just like, wow, this is mind-blowing. And so the beauty of it is that we can now take, for instance, patients that have a genetically caused disease and create organoids from them. And we can take normal, healthy, matched controls and see what makes the little brains, say, from the patients. How do they look different from the little brains from the controls?

And now by comparing and contrasting, you say, okay, this is what sick looks like. This is what healthy looks like. Now let's find something that reverts the sick to healthy. And that's exactly where the machine learning comes in because you can't, I mean, people have a hard time just by looking at the little brains. I don't even know what to look for. But Christian learning doesn't have a problem with that. So that's actually a thing that we do,

and it's the core of a lot of what we do it in C.TRO. What do you think about our lifespan and the ability to have a healthy life for a longer period of time? Are we going to increase our lifespan dramatically for the people who hear our voice right now, whether it's a 15-year-old or a 50-year-old? or are we just going to get healthier and healthier so that 70-year-olds are going to be an 80-year-olds are going to be skiing because they just have a better, they'll be in better shape because our bodies won't degrade as much

and we'll be able to treat so many diseases, whether it's a bad knee or a bad liver. Honestly, I don't think any of us really know the answer to that. There have been experiments in model animals, worms, to a lesser extent, mice that have extended the lifespan dramatically. I don't know if anyone has demonstrated conclusively that that same lifespan extension is even possible in humans. And I don't think we know if there is a kind of natural end to how much a cell can continue to propagate and live before it just incurs enough wear and tear that it just kind of, you know, stops. And I think that's one hypothesis. No one has There is no feasibility proof of infinite life, I think, for, or, you know, considerable life extension.

So I, but honestly, from my perspective, if we're able to get it to the point that you and I live to 120 and then we go to sleep one night and then we don't wake up, that is so much better than the old, you know, degeneration and decrepitude that you see for a lot of people as they age, I would frankly be delighted. delighted if we were able to do that. That would be the ideal situation. Heck, if I could just live to 90 or 100 and just be able to still ski or pick up my daughters or their grandkids or their grandkids, and at the pace we're going in terms of podcasts, like once every seven years, that would get us 10 more podcasts under our mouths to continue this discussion. When you see what's happening in biology, when you see what's happening in machine learning. I don't know if you saw the GPT3 thing that OpenAI launched. last week where they can sort of fill in text.

You put text in. It tells you what the next text. Do you see that? The little Open AI project? I saw the previous version. I haven't seen the most recent one. People are kind of losing their minds over it. It's sort of for people listening.

It's like when Gmail tells you the next three words that you're probably going to say and it just feels like, whoa. Obviously, that is a kind of a preview of the work you're doing, which is you're going to start to have the signs tell you, hey, here's the next two or three words, two or three genes, two or three drugs, molecules, whatever, to work on. When you start seeing this, does it make you believe that all of this, and by this I mean our lives, our consciousness, the big C question, is this all some organic, you know, process that's occurring?

Are we in a simulation? Or do you believe in God? When you look at this and you think to yourself in your private moments like, this is too complex to be random. There has to be a god. Or did you just think, oh, this is so predictable, and we're unpacking it at such a velocity that we're just understanding, you know,

on a scientific basis, biology at such a great level. Do you believe in, like, a higher being? I'm curious what this job leads one to believe. So I don't know that I'm representative of others, but to me, these are actually two separate questions and one almost doesn't speak to the other. So the ability that we have been able to create machine learning algorithms that are really, really good at pattern matching with given enough data,

to me is an amazing engineering feat. And I'm deeply impressed that we were able to get there. But those algorithms understand nothing. Right. They are really good at having seen enough, sequences of words that lead to the next three words that they're really good at making that prediction. And that's awesome and potentially super useful.

But you take those algorithms out of their comfort zone or the place that they were designed to function, you put them somewhere else. They have none of that adaptability that we as humans have to grok a situation, figure out what's going on. And think of the Martian as an example of a movie, right? You took a person and put him in a completely different situation. situation and he just figured it out. And that is a thing that we have not been able to build into our machine learning models. And furthermore, the techniques for doing so, I haven't

seen anything that I've found as being on the path to creating that level of flexible thinking and adaptability. So I think it's great that there's these engineering efforts out there. It doesn't speak to me to the existence of God. And, you know, I can certainly speak about my personal beliefs about God or not. So putting the machine stuff aside, like we're primates that have made these incredible tools, right? We were tool builders for so long.

But I'm talking about when you start looking at the complexity of biology now, because that is a separate thing, right? It's like there's the tools, and then there's what the tools are uncovering. And since you brought up the Martian, I don't know if you ever saw the movie Prometheus, you know, the sequel to aliens.

Did you ever see it? No. Well, you have to see it because it really tries to answer the question of like, you know the movie Alien, of course. Of course. So in that movie, there were engineers, theoretically, that built the aliens as some sort of weapon or experiment.

It was never answered. And then Prometheus is really Scott's follow-up where he tries to answer that question where people are doing biological experiments. And it does make wonder, at least it makes me wonder, like, this is so complex and so beautiful, right? Does it, it almost makes an atheist like. me believe like there's some higher power that constructed this. How did this all start when, you know, like you start looking at it? I'm taking a second shot at your thoughts on.

Well, I'm happy to, I see the amazing power of evolution. Now, whether evolution is derived from a higher being or not, I guess I would say I'm an agnostic because I believe in data and not having data to speak to the existence or lack thereof of a higher being. I basically reserve judgment, but evolution is an unbelievable force in the creation of incredibly beautiful and complex systems that just do amazing things. And I find in some ways, machine learning is kind of that too. It's the power of progressive evolution and improvement as getting us to places that we would never otherwise anticipate. And I think that is a miracle in itself, regardless of whether it comes from a higher being or not. It's incremental improvement just over time that compounding leads to something absolutely just astounding.

Miraculous. In fact, yes. All right, listen, I'm assuming you're hiring the smartest people in the smartest people in the world to try to solve these problems. So who are you looking for? And if there's somebody in college who's just crushing it or taking online courses on Coursera and they're just genius level, what are you looking for out there to join in CITRO?

So we love people who are broad-minded and we think of them as bilingual or multilingual. So people who simultaneously speak and love biology and speak and love data and Obviously, they don't necessarily have to be experts in both, but the willingness to sort of engage with a discipline that is not your own and think creatively about that with someone who comes in from that other discipline who's also there with you to kind of jointly brainstorm, those are the people that we love, the ones that are really willing to come in with humility and say there's just so much, I mean, I have my expertise, but there's so, so much out there that I don't understand. And together, as a team, we will arrive at heights. We will be able to create things that we would never be able to do on our own. I think that's really the kind of person that we look for. The people who want to be part of a team that's aspiring to something really amazing where the whole is going to be considerably larger than the sum of the parts. It's that range and that interdisciplinary skill that creates breakthrough, correct?

Absolutely. But it's so hard because academics and the world wants specialization. So we push people toward specializing and computer sciences and AI is over here and then biology and is over here and physics is over here and finances over here. It's when people actually can cross over over that maybe breakthroughs occur, yeah? I agree. And I think it's, uh, what's also critical is to recognize that you as an individual are never going to be the expert. on everything that's necessary for whatever it is that we're trying to build. And that's why the willingness to engage in constructive and respectful dialogue with others is really an

important piece of it because we are much, much stronger as a team than we are as individuals. And that is that cultural piece, that mindset piece is as one's technical skills as we hire people at Enicero. Yeah, you have to be able to sit around that table and brainstorm and make those breakthroughs. You can't think that you know it all just because you're the top of your field because the breakthroughs come from the crossover. So listen, Definitely, it's great to speak to you again. Congratulations on the progress. We're all rooting for you.

And if you're a really brilliant person or you know a really brilliant person in these areas and you want to be a drug hunter and you want to save lives and reduce suffering in the world. You know, definitely's been out this for a while as an entrepreneur. and a professor. Can't think of a better place for you to put your brains. We do not need you to go make more people click on ads in social networks. For the love of God, if you're smart. Do not go to Wall Street. Do not go to Facebook and make people trend, garbage stories.

Go work for Encitro and try to save some lives. It's better. Right? I mean, isn't this one of the problems that you see is like the financial incentive to go do stupid smart people to go do stupid inane work is too great in the world. I couldn't agree with you more. I think if more of the really smart people that are out there said, what can I do to make the world a better place? I think we would be in a much better shape as a society. I try and teach that to my children and I tell them, look, for those of you who were

born to a privileged situation, in the sense that you have a good home and get to go to a good school, you should be looking to give something back. You should be looking at how you can leave the world a better place than the one you came into by virtue of something that you've done. So I completely agree with you. I think for those of you who are smart and will have that opportunity, give something back to the world. Make a dent in the universe to paraphrase Steve Jobs. Yeah, make a dent and don't like make a spreadsheet. Oh, God.

All right. Thanks, Stephanie. Stay safe and we'll see you all next time on this week's service. Bye-bye.