Rob Reid on biosecurity risks: synthetic biology is exponential, and DNA supply chains are dangerously unregulated

Speaker 2: You've made it. You're worth taking a shot at. Yeah. Like your people are talking about you. You're not a nobody anymore. Anyway, we have Rob Reid in the waiting room. Let's bring him in to the TBPN UltraGone. Rob, how are you doing? Good now. Hello, gentlemen. Thank you so much What's going on? Taking the time. Great to have you. We wanted to talk about all the advancements in bio, biosecurity, what's happening there. But since it's the first time on the show, I would love to know a little bit about your background and sort of introduce you to the audience. Could you take us through some of your earlier journeys, some of the companies that you've worked with, and what you're doing now? And then we can go into the hot topics of today.

Speaker 7: Absolutely. So I came to The Valley right out of school, and I'm chronologically advanced enough that it was during the first wave of the Internet boom. So I was at a big company that got very aware of the Internet's sort of oncoming tide early, I'd say. Got into the Internet group there. Became an Internet person, you know, in, like, 1994, which is very, very early. And as a result of that, I saw a lot of things coming. I ended up writing a book about the rise of the Internet as a commercial medium, kind of as it was happening. And that got me well widely known that I became sort of a junior VC at a junior firm. Mhmm. And then I took off And what

Speaker 1: What year did you publish the book? Sorry to

Speaker 6: interrupt.

Speaker 7: The book came out in 1997. It's called Architects of the Web. And I profiled Yeah. I this is

Speaker 1: I think this think be a good business opportunity. There there we there might be some commerce that happens on the World Wide Web.

Speaker 2: That's amazing. Yeah.

Speaker 7: Yeah. That was one of my theories. There might be commerce. There people might trade stocks.

Speaker 4: Yeah.

Speaker 7: Video. Crazy stuff.

Speaker 1: How much pushback did you get at the because because we we we've covered it before on the show, there was this almost if you look at the doom and fear that people had around the Internet, it almost aligns one to one with with AI. There was

Speaker 2: The y two k moment.

Speaker 1: Yeah. Y two k cyber.

Speaker 2: Putting all bookstores, all retail, the mall apocalypse, retail Yeah.

Speaker 1: The past apocalypse was was the retail apocalypse.

Speaker 2: Yeah.

Speaker 1: Almost there's there's basically a one to one example for each. Yeah.

Speaker 7: This was in the honeymoon period before the panicking started. So '97, it's still, you know, primarily a tech forward thing. Mhmm. And there was wild optimism about what was gonna happen. Mhmm. Although we didn't have the vocabulary yet, I think if you ask people what the Internet would do for information and informedness, people would describe a combination of Wikipedia and Snopes, if you remember Snopes.

Speaker 2: Oh, yeah.

Speaker 7: Nobody ever imagined that disinformation would spread rapidly, that it would be divisive. It was a real Kumbaya vibe early on, and people were obviously a little bit overoptimistic on that front.

Speaker 2: Yeah.

Speaker 7: But boy, were they right about the economic opportunity and the degree to which it would change our lives and and the way we do business. And so at that time with that book, you know, it was it was celebrating that moment and really taking a deep deep look at what was happening right then at the outset. And it was a small enough industry, and I had been a part of it just a few months before I started writing it, that I had access to everybody, and I got to know everybody of any consequence. I mean, this is when, you know, you could go over to Netscape and get a meeting with Marc Andreessen if you had a good reason to with relatively little difficulty. And through all the relationships that I I I made through that, I ended up getting this sort of field promotion to be a VC far sooner than I I thought I would. Did that for a couple years, learned a ton, started this music I started a company as a solo founder of a a company that created a a music service called Rhapsody.

Speaker 2: But tell us the domain name of that company.

Speaker 7: Oh, listen.com.

Speaker 2: Fantastic domain name. We love these amazing domains. They're so rare these days. But back then, everyone had a great domain. What what was the entrepreneurial journey like? Was it a tough transition? What was the whole, journey like? How do you reflect on it today?

Speaker 7: You know, the toughest thing was that I'd never really managed prior to that. I'd been a mid level marketing grind at a big computer company and then a VC. Learned a lot of lessons as a VC, but had to be a self taught manager.

Speaker 1: Mhmm.

Speaker 7: And that's hard enough when you're doing it for the first time. But when you have people who work for people who work for people who work for you, you know, that's a lot to to rocket through. We got up to about 200 employees before we sold. And so that was the trickiest thing. And it is lonely being a solo founder because there's no one person. You may have great investor relationships, great board relationships, you know, even great relationships, obviously, with your team. There's nobody that looks at things through precisely the same lens as you. Mhmm. That said, every cofounder that's added to an equation or every cofounder relationship, which which scales quadratically if you go to three, four, etcetera, is an access upon which a startup can break. Yeah. And so that's that's sort of the trade off that you face. And, you know, I'm glad I did it as a solo founder for a whole bunch of reasons, but that's tough.

Speaker 2: Yeah. Interesting. When did you start thinking about bio risk? Because generally, that seems like a second order effect of technology at this point. But what piqued your interest? What was your process to get up to speed and start actually thinking through tactical implications and forecasting risk?

Speaker 7: In a way, it connects to that startup Rhapsody. So most of your your listeners may not remember Rhapsody. But for all intents and purposes, we were the first Spotify. We created the unlimited on demand streaming model that everybody else later adopted. We were the first company to sign, you know, full catalog licenses with all the major labels even before Apple did for their m p three start. And when I came out of that, I wanted to take a little bit of time off. And I wrote I I'd been I'd been a writer. I'd written a couple of books, but nonfiction. And so I wrote a book about a crazy alien civilization that was so into American pop music, they accidentally committed the biggest copyright infringement in human in the history of the universe. The gantry. The entire universe. Yeah. Based based on a true story, I like to say. And so that I was I thought I was doing for myself. Bizarrely enough, it it briefly and barely, I wanna put those words out there, became a New York Times bestseller. And I was living in LA forever.

Speaker 1: I know. You said briefly, but it's forever.

Speaker 2: Yeah. Yeah. It's like I

Speaker 7: was like a little prairie dog coming up to the bottom of the list and looking around for a couple weeks, but you you do get to say that.

Speaker 1: Forever.

Speaker 7: And and I was living in Los Angeles with we were just getting married. My my wife was hosting a TV show that kinda had us marooned down there for a long period of time. So before starting another company, I leaned into more writing and doing a bunch of deep science podcasting, and this is where it all sort of comes together. Often I mean, generally, interviewing world class scientists all the way up to the Nobel level about, you know, very deeply about their work. And I would spend thirty or forty hours of prep before each interview because I was mainly doing for the fun of it. That was a great way to learn about a bunch of different things. Through one of my novels and then through my podcasting, I got more and more concerned about synthetic biology risk. And, you know, this probably started about ten years ago. And at the time, you could see that synthetic biology, like computing, was and is an exponential technology, which means that the things that are incredibly hard or indeed even impossible to do today, unless you're one of two or three people, are the types of things that freshmen in college will be able to do with the passage of time. And when you think about the bad things that can happen on the edges of synthetic biology, and this is before we got to the AI intersection, you could really imagine catastrophic acts of bioterror or bioerror, you know, seven, eight years in the future. And, you know, I started diving deeper into this, speaking about it, writing about it. And then right before COVID, I gave a TED talk about the dangers of an unnatural pandemic. And, of course, a few months later, we had COVID. And so I've gotten deeper and deeper into this. It's my public service side. I have made one investment that is connected to this as a VC, which is what I do now. But it's my public service side, and I've, at this point, probably put, you know, at least a couple thousand hours into it over the years.

Speaker 2: There's there's been some movement, at least, like open letter signed by AI lab leaders. The most recent was around basically record keeping and sort of KYC rules that companies that manufacture DNA, RNA sequences, if I understand it correctly. Is that the right hinge point to focus on? Are you a fan of an ensemble approach where the models have guardrails baked in? And then also, if you're using the model through a SaaS product, then the software also is tracking things. Then the company that makes the actual sequence might have guardrails, and then maybe the equipment is regulated with licenses. Do you have a prescriptive approach that you're advocating for at this point? Is it evolving? How are you thinking about that?

Speaker 7: It's always evolving, but there are a few very, very simple and incredibly inexpensive things that we can do as first steps that would significantly improve the situation. So first of all, all of the foundation model companies are very, very aware of this risk. They all have what's called classifier models that are there. We've all probably experienced versions of them

Speaker 2: Yeah.

Speaker 7: That are there to prevent people from doing, you know

Speaker 1: Yeah.

Speaker 7: Dangerous biological research or try to create dangerous bugs. Every frontier model, every model has what's called a system card, which they describe the model and its capabilities. And every system card I've ever seen has acknowledged that classifiers classifier models, which are like smaller models, almost like bouncers that sit in front of the big model. Classifiers can and are always fooled. So they are doing everything they can. They will continue to. We should be grateful for that, but we know that that's only the beginning of it. The next issue is securing our DNA supply. So at this point, practically anybody who wants to make a long strand of DNA goes to a service bureau to have that done because the experts are better at it Mhmm. And researchers would rather do research. There is a a consortium called the International Genome Synthesis Sequencing Consortium or something like that. It's called the IGSC. Mhmm. It's a voluntary consortium that does very good work. They put together what they call sequences of concern, and if somebody tries to order one of those things, you know, traffic light red yellow will go off, and they'll dive into the order order. This is expensive for companies to comply with, and it's also voluntary. The IGSC complaint claims that they speak for 80% of the synthesized DNA in the world, but that was kind of a out of the, you know you know, kind of blue sky guess that they made over ten years ago. And they don't have very many Chinese members, so that's no longer true. But even if it was 80%, I I can confidently say that if there were a town in which 80% of the bars enforce the drinking age, that would be a town without a drinking age. Right? Mhmm. So it does nothing. A bad actor is going to go to a non IGSC shop, obviously, and it's even worse than that. It'd be almost like you had security lines at the airport that were voluntary, and it was actually cheaper to fly if you went if you didn't go through the security line because you save money by not you know, these this adds to the cost basis. We need to make that mandatory.

Speaker 2: Mhmm.

Speaker 7: And there's actually an open letter signed by a lot of folks, including from AI Labs, just a few days ago pushing for that. Yeah. This is a dead simple thing. And by the way, it's as much in China's interests as it is in our own. Because, you know, if anybody can order any stranded DNA that they want, and we'll get into what they can do with that DNA once they have it, if you'd like, the the two entities that have the most to lose are The United States and China. And this is something that we can team up on with them and say, yeah. There's gonna there needs to be an international regime. It has to be mandatory, and it also has to apply to benchtop synthesizers where people can make much shorter strands without having to go to a service bureau.

Speaker 1: Mhmm. Bad actor gets a hold of or or is working with with somebody that's not voluntarily being, I guess, regulated, what can they do with it?

Speaker 7: Well, in a worst case scenario, let's think about something that's not only not just hypothetical. Let's say that there is a really deadly genome that's in wide circulation online. Everybody knows what it is. Nineteen eighteen flu happens to be just such a genome because our government put it there. They put it there a long time ago when very few people could do anything with it, but the passage of time always happens. And it was a very foolish thing to do because we now live in a world in which 30,000 people roughly, perhaps a little more, have the tools and the know how, which would allow them to basically animate that virus just from the DNA if they felt like doing it. And the last time that thing came around, it killed a staggeringly high percentage of the people in the world, and we don't want that to happen again. Mhmm. There's reasons to think it might not turn into a pandemic this time, but let's skip the rabbit hole. This is just a hypothetical. There is a brilliant professor at MIT named Kevin Esvelt who stress tested the system with this in mind, and he basically cut up a whole bunch of orders which collectively added up to Spanish flu and sent it to a whole bunch of different labs. Most of those labs actually knew that this was a sequence that was part of the Spanish flu. But although the Spanish flu virus is very, very tightly regulated by our government, it's one of the very most regulated biological, you know, critters or sequences or whatever you wanna say in the world because it's called a select agent. There's fewer than a 100 of those. They didn't stop themselves from sending the DNA because it's only the entire genome that's regulated. And so this was a scientific paper he did. He did it with the knowledge and cooperation of the FBI. It was a red teaming exercise. Mhmm. And even going to IGSC compliant labs, he could pull this all together, and somebody like Kevin and about 30,000 other people could then animate the virus. And the number of people who can do that is only gonna go up. It was probably fewer than 30 when that genome was first posted to the Internet. At some point, it will be anybody in freshman biology. Mhmm.

Speaker 1: What So please. What do you I feel like every six months, there's a very concerning headline where like some illegal bio lab is like raided by law enforcement and then it just sort of like goes away. What do you make of that? I'm sure you've kind of tracked a lot of these stories.

Speaker 2: Yeah.

Speaker 1: How I'm assuming if law enforcement is consistently finding them that there's a lot more out there that aren't found. Why do people why are they created in the first place? What is The United States doing to crack down further? All that stuff.

Speaker 7: Yeah. I mean, we can rely on the fact that we have not found each and every one of these things. And this all keys off of proliferation. So, again, I'll go back to the notion that synthetic biology is an exponential technology like computing. With something like that, you have very, very, very rapid proliferation and very, very, very rapid drops in cost. So just think about transistors. I I forget the statistic, but was in the late fifties that one very brilliant person in the entire world developed the ability to play checkers with a computer. You know, he had one of five machines that could do it. He was one of two people who were smart enough to program or whatever. And with the passage of a couple decades, you know, anybody can play chess, and then it gets to the point that any, you know, street kid with a tiny budget can do that. We'll go through the same thing with Synbio. Now these labs that we're discovering have not been terribly capable yet, But because particularly with when Synbio meets AI, the level of expertise that you need in order to achieve really powerful things that perhaps nobody even could achieve right now will become increasingly cheap achievable in a very, very low budget with the passage of, like, a single digit number of years. And so these labs are we're hearing more and more about them because it's becoming cheaper and easier to create them. And, you know, this needs to be something that we as a society think about and invest in as much as cybersecurity. We spend tens of billions of dollars a year on that as we should. Mhmm. But the the consequences of, you know, the human genome and therefore the human species being taken down by a cunningly engineered virus are so much higher than that in cyber that we should be rationally making at least as much of an investment. And at this point, we're making a vanishingly small amount of investment. And I I I should point out, part of this investment needs to be about, you know, preventing bad actors without hobbling our ability to get incredible gains from Synbio and from AI models. And so that has to be done very deftly and very intelligently. And so simply passing a law and saying we've passed the law ain't enough. Not that we're in any danger of passing a law that makes us more secure on a bio level, but the the the objective, the urgent objective needs to be to allow creativity and and technological and scientific advance to run rapid and bring incredible benefits to us, but not allowing ourselves to get taken down. And this thing that I mentioned with SecureDNA, that would cost, you know, virtually nothing because there are tools that do this that are actually open source. It's just not required to anybody. There just needs to be a requirement in that case in some enforcement, and we haven't yet had the social willpower to even do that. Hopefully, that's gonna change with all these powerful signatures that are starting to accrue on these open letters.

Speaker 2: Seems

Speaker 1: pretty straightforward that AI's ability to find software vulnerabilities, you know, basically sending a basically sending a model and saying, hey, just figure out vulnerabilities in, you know, this infrastructure or this piece of software. Same do do you believe, like, already the models have that same ability to effectively find, like, vulnerabilities in Yeah. The human

Speaker 7: There there's there's no question because we are not as robust as the macOS, for instance. There are lots and lots of known vulnerabilities that we have already that have not been patched, will never be patched, cannot be patched. And so, yeah, we're we're a much softer target, you know, than Linux or macOS.

Speaker 1: And so how does how does it seems like the solution is is needs to be like you've been talking about focused on like actually making it more difficult to procure DNA. Because if you assume that open models, you know, a bad actor is gonna be able to have an open model with no guardrails, with access to enough compute that they can send the model on, you know, basically a goal which is like find as many possible, you know, known vulnerabilities as well as potentially net new vulnerabilities in, you know, the the human species. So that it feels like cat's out of the bag there maybe already or will be within the next twelve months. But do you think

Speaker 7: Hopefully, '24 or '36, but perhaps already.

Speaker 2: Sure. Yeah. Yeah. Last question for me. You know, as as as guns proliferated society, so did metal detectors. Is there an opportunity for, more diffuse early warning detection around sickness? I'm just thinking if I'm if a WHOOP or an Oura Ring or some other connected device can tell every individual, you're getting sick. You should get checked out. You shouldn't go to work. You shouldn't go to that populate you shouldn't spread anything. Feels like that could make quarantining more targeted and potentially more of an early warning detection system? Are you optimistic about that?

Speaker 7: You've actually hit on a couple of really interesting things with that question. Mhmm. Let's go to metal detectors and guns first. In 1972, astounding fact, there were over 50 hijackings of large commercial airliners in The United States. Yeah. That started being true in the late sixties. Yeah. Why? There was literally no airport security at all in The United States. And who was lobbying against it? The airlines.

Speaker 2: Oh.

Speaker 7: Because they thought it would be expensive and it would make people think seeing the security gates would make people think that airline travel was dangerous. Yeah. I'll tell you what would make me think it's dangerous, 50 hijackings

Speaker 2: a year.

Speaker 7: So and Hypers. I'm sorry. Go ahead.

Speaker 1: That people would hijack a plane and enjoy Hostages.

Speaker 4: Hostages. Save me money.

Speaker 1: So there was there a lot of like ransom payments being made? Was that like the primary

Speaker 2: gotta do a whole deep dive op here.

Speaker 1: I guess we're gonna do three hours tomorrow on on The

Speaker 2: hijack There

Speaker 7: was everything. There was everything. So there there's a great book called The Skies Belong to Us to tell the story, but this was kind of the late sixties, and there were a lot of radicals, and the most popular destination was Cuba. Somebody would put it in their heads that they're gonna fly down to Cuba with a hijacked plane. Fidel's gonna become their best friend, and it's just gonna be, you know, rum coladas and beaches for the rest of their life. It literally got so bad that they started a a dormitory for wayward American hostages, and I'm not exaggerating, in Havana. And they would send the plane back and just have to deal with these, you know, entitled, you know, college or post college kids. Anyway, the security guards take gates go in, and we go from 50 hijackings in '72 to zero in '73. Yeah. And then there's kind of a very occasional thing, and then 09:11 hits. Yeah. And after nine eleven hits, we take security much more seriously than ever, and there has not been a single one in The United States in, you know, twenty five years.

Speaker 6: Yep.

Speaker 7: What does this tell us? It means that you just you don't wanna let the perfect be the enemy of the good. Mhmm. A lot of people say, well, gee, securing the DNA is only part of the solution. It is. But you make a major speed bump in front of people. It becomes a much higher bar, much higher intent. Something you don't do on a whim because you're pissed off that your girlfriend or boyfriend broke up with you. And then it becomes much, you know, more difficult to get over that bar, and you can really, really plummet the level of ambient risk down to the point that it's really only determined actors who are out there who become much easier to detect using intelligence and law enforcement and all the things that that we wanna use. And then in terms of your question of surveillance, yeah. We're pretty far off from when Oura Rings could necessarily do that, but we should be thinking about improving them and tying them into the primary care system in The US, if only because we're worried about sick days at work. Like, there's lots of benefits to that. But the other thing we can be doing is what's called biosurveillance

Speaker 5: Mhmm.

Speaker 7: Which is getting a global detection system out there. The best place, believe it or not, is airport airports

Speaker 2: Yeah.

Speaker 7: Where everybody comes in and all the waste products from all of those planes goes into a single place where you can sequence parts of that wastewater and see if there is a genetic sequence that's never been seen before that is suddenly starting to grow exponentially in various points of the world. Because that would mean it's either something with a long, incubation cycle, or it's what some people call a stealth pandemic, which is something that is engineered to have an extravagantly long incubation cycle. So everybody's asymptomatic, but it's spreading and spreading. If that ever happens, we wanna know pronto. And the budget for some very, very powerful plans for biosurveillance is is been as low as, you know, $50,000,000 set up and 10,000,000 a year thereafter. These are not expensive things.

Speaker 2: Yeah. Yeah. We gotta do it. Well, thank you so much for taking the time. We really

Speaker 1: Let's do this again soon.

Speaker 2: We got do this again soon. This is fantastic. Also, there's a bunch of your fans in the chat hoping for more podcasts from you because they

Speaker 1: Your podcast.

Speaker 2: Yes. From your

Speaker 7: I'll I'll talk to

Speaker 2: them. Great. Well, thank you so much. Rob. Great to have great of your day. We'll talk to you soon. Let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, mobile, on social, on marketplaces, and now with AI agents. And without further ado, we have the founder and CEO of Taste Labs. We talked about it yesterday on the show.