Lex Fridman PodcastRob Reid: The Existential Threat of Engineered Viruses and Lab Leaks | Lex Fridman Podcast #193
EVERY SPOKEN WORD
150 min read · 30,009 words- 0:00 – 2:28
Introduction
- LFLex Fridman
The following is a conversation with Rob Reid, entrepreneur, author, and host of the After On podcast. Sam Harris recommended that I absolutely must talk to Rob about his recent work on the future of engineered pandemics. I then listened to the four-hour special episode of Sam's Making Sense podcast with Rob, titled Engineering the Apocalypse, and I was floored and knew I had to talk to him. Quick mention of our sponsors: Athletic Greens, Belcampo, Fundrise, and NetSuite. Check them out in the description to support this podcast. As a side note, let me say a few words about the lab leak hypothesis, which proposes that COVID-19 is a product of gain-of-function research on coronaviruses conducted at the Wuhan Institute of Virology that was then accidentally leaked due to human error. For context, this lab is Biosafety Level 4, BSL-4, and it investigates coronaviruses. BSL-4 is the highest level of safety, but if you look at all the human and the loop pieces required to achieve this level of safety, it becomes clear that even BSL-4 labs are highly susceptible to human error. To me, whether the virus leaked from the lab or not, getting to the bottom of what happened is about much more than this particular catastrophic case. It is a test for our scientific, political, journalistic, and social institutions of how well we can prepare and respond to threats that can cripple or destroy human civilization. If we continue gain-of-function research on viruses, eventually these viruses will leak, and they will be more deadly and more contagious. We can pretend that won't happen, or we can openly and honestly talk about the risks involved. This research can both save and destroy human life on Earth as we know it. It's a powerful double-edged sword. If YouTube and other platforms censor conversations about this, if scientists self-censor conversations about this, we'll become merely victims of our brief homo sapiens story, not its heroes. As I said before, too carelessly labeling ideas as misinformation and dismissing them because of that will eventually destroy our ability to discover the truth. And without truth, we don't have a fighting chance against the Great Filter before us. This is the Lex Fridman Podcast, and here is my conversation with Rob Reid.
- 2:28 – 8:47
The most entertaining outcome is the most likely
- LFLex Fridman
I have seen evidence on the internet that you have a sense of humor, allegedly, but you also talk and think about the destruction of human civilization. What do you think of the Elon Musk hypothesis that the most entertaining outcome is the most likely? And he, I think, followed on to say as seen (laughs) from a, an external observer, like if somebody was watching us, it seems we come up with creative ways of, uh, progressing our civilization that's fun to watch.
- RRRob Reid
Yeah, so he, he... exactly. He said from the standpoint of the observer, not the participant, I think.
- LFLex Fridman
Right.
- RRRob Reid
And so what's interesting about that, those were, I think, just a couple of freestanding tweets and, and delivered without a whole lot of wrapper of context so it's left to the mind of the, the reader of the tweets-
- LFLex Fridman
Yes.
- RRRob Reid
... to infer what he was talking about. But so that's kind of like... it provokes some interesting thoughts. Like first of all, it presupposes the existence of an observer, and it also presupposes that the observer wishes to be entertained and has some mechanism of enforcing their desire to be entertained. So there's like a lot underpinning that. And to me that suggests, particularly coming from Elon, that it's a reference to simulation theory-
- LFLex Fridman
(laughs)
- RRRob Reid
... that, you know, somebody is out there and has far greater insights and a far greater ability to, let's say, peer into a single individual life and find that entertaining and full of plot twists and surprises and either a happy or tragic ending, or they have a incredible meta view and they can watch the arc of civilization unfolding in a way that is entertaining and full of plot twists and surprises and a happy or unhappy ending. So okay, so we're presupposing an observer. Then on top of that, when you think about it, you're also presupposing a producer because the act of observation is mostly fun if there are plot twists and surprises and other developments that you weren't foreseeing. I have re-read my own novels, and that's fun because it's something that I worked hard on and I slaved over and I love, but there aren't a lot of surprises in there. So now I'm thinking we need a producer and an observer for that to be true. And on top of that, it's got to be a very competent producer because Elon said the most entertaining outcome is the most likely one. So there's lots of layers for thinking about that. And when you've got a producer who's trying to make it entertaining, it makes me think of there was a South Park episode in which Earth turned out to be a reality show-
- LFLex Fridman
Yeah.
- RRRob Reid
... and somehow we had failed to entertain the audience as much as we used to so the Earth show was going to get canceled, et cetera. Um, so taking all that together, and I'm obviously being a little bit playful in laying this out, what is the evidence that we have that there... we are in a reality that is intended to be most entertaining? Now, you could look at that reality on the level of individual lives or the whole arc of civilization, other lives, you know, levels as well, I'm sure. But just looking from my own life, I think I'd make a pretty lousy show. I spend an inordinate ama- amount of time just looking at a computer. I don't think that's very entertaining. And there's just a completely inadequate level of shootouts and car chases in my life.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
I mean, I'll go weeks, even months without a single shootout or car chase.
- LFLex Fridman
That just means that you're one of the non-player characters in this game. You're just waiting to make-
- RRRob Reid
I'm an extra.
- LFLex Fridman
You're an extra that waiting for your one opportunity for a brief moment to actually interact with one of the main, um, one of the main characters in the play.
- RRRob Reid
Very interesting. Okay, that's, that's good. So okay, so we rule out me being the star of the show, which I probably could have guessed that anyway, but then even the arc of civilization.
- LFLex Fridman
Yeah.
- RRRob Reid
I mean, there have been a lot of really intriguing things that have happened and a lot of astounding things that have happened, but, you know, I would have some werewolves, I'd have some zombies, you know, I would have some really improbable developments like maybe Canada absorbing the United States, you know, so I don't know. I'm not sure if we're necessarily designed for maximum entertainment, but if we are, uh, that will mean that 2020 is just a prequel for even more bizarre years ahead. So I kind of hope that we're not designed for maximum entertainment.
- LFLex Fridman
(laughs) Well, the night is still young in terms of Canada, but do you think it's possible for the observer and the producer to be kind of emergent? So-... meaning, it does seem when you kind of watch memes on the internet-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... the funny ones, the entertaining ones spread more efficiently.
- RRRob Reid
They do.
- LFLex Fridman
I mean, I don't know what it is about the human mind that soaks up en masse funny things m- m- much more sort of aggressively. It's more viral, like, in, in the full sense of that word. Is, is there some sense that whatever this, the evolutionary process that created our cognitive capabilities is the same process that's going to, in an emergent way-
- RRRob Reid
Mmm.
- LFLex Fridman
... create the most entertaining outcome, the most meme-ifiable outcome, the most viral outcome if we were to share it on Twitter?
- RRRob Reid
Yeah, that's interesting. Um, yeah, we do have an incredible ability, like, I mean, how many memes are created in a given day? And the ones that go viral are almost uniformly funny, at least to somebody with a particular sense of humor.
- LFLex Fridman
Right.
- RRRob Reid
Um, yeah, I'd have to think about that. Um, we are definitely great at creating atomized units of funny.
- LFLex Fridman
Yeah.
- RRRob Reid
Like in, in the e- example that you used, there are going to be X million brains parsing and judging whether this meme is retweetable or not.
- 8:47 – 12:07
Meme theory
- RRRob Reid
- LFLex Fridman
Do you find it compelling or useful to think about human civilization from the, uh, perspective of the ideas versus the perspective of the individual human brains?
- RRRob Reid
Mmm.
- LFLex Fridman
So, almost thinking about the ideas or the memes, this is the Dawkins thing-
- RRRob Reid
Mmm.
- LFLex Fridman
... as the organisms.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
And then the humans as just, like, uh, vehicles for briefly carrying those organisms as they jump around and spread.
- RRRob Reid
Yeah, for propagating them, mutating them, putting selective pressure on them-
- LFLex Fridman
Yeah.
- RRRob Reid
... et cetera.
- LFLex Fridman
Yeah.
- RRRob Reid
Um, I mean, I found, uh, Dawkins' interpret, or his, his launching of the idea of memes is just kind of an afterthought to his unbelievably brilliant book about The Selfish Gene. Like, what a P.S. to put at the end (laughs) of a long chunk of writing.
- LFLex Fridman
(laughs)
- RRRob Reid
Profoundly interesting. I view the relationship though between human and, humans and memes is probably an oversimplification, but maybe a little bit like the relationship between flowers and bees, right?
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Do flowers have bees or do v- bees, in a sense, have flowers? And the answer is, it, it is a very, very symbiotic relationship in which both have semi-independent roles that they play and both are highly dependent upon the other.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And so, in the case of bees obviously, you know, you could see the flower as being this monolithic structure physically in relation to any given bee, and it's the source of food and sustenance, so you could kind of say, "Well, flowers have bees." But on the other hand, the flowers would obviously be doomed if they weren't being pollinated by the bees, so you could kind of say, "Well, you know, bees are, you know, flowers are really expression of what the bees need." And the truth is a symbiosis. So with, with memes and human minds, our brains are, are clearly the petri dishes in which memes are either propagated or not propagated, get mutated or don't get mutated, it's they are the venue in which competition, selective competition plays out between different memes. So, all of that is very true and you could look at that and say, really, the human mind is a production of memes, and ideas have us rather than us having ideas. But at the same time, let's take, um, a catchy tune as an example of a meme. Um, that catchy tune did originate in a human mind. Somebody had to structure that thing. And as much as I like Elizabeth Gilbert's TED Talk about how the universe, um, I'm simplifying, but you know, kind of the ideas find their way, in this beautiful TED Talk, it's very lyrical, she talked about, you know, ideas and prose kind of beaming into our minds and, you know, she talked about needing to pull over to the side of the road when she got inspiration for a particular paragraph or a particular idea and a burning need to write that down. Um, I love that, I find that beautiful. As a, as a writer, as a novelist, uh, myself, I've never had that (laughs) experience. And, uh, I think that really most things that do become memes are the product of a great deal of deliberate and willful exertion of a conscious mind. And so, like the bees and the flowers, I think there's a great symbiosis, and they both kind of have one another. Ideas have us but we have ideas
- 12:07 – 18:54
Writing process
- RRRob Reid
for real.
- LFLex Fridman
If we could take a, a little bit of a tangent-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... with Stephen King on writing, you as a great writer, you, you're dropping a hint here that the ideas don't come to you, they, (laughs) that it's a grind of sort of, uh, it's almost like you're mining for gold. It's more of a very, uh, deliberate, rigorous daily process. So maybe, can you talk about the writing process? How do you write well? And, and maybe if you want to step outside of yourself, almost like give advice to a, an aspiring writer, what does it take to write-
- RRRob Reid
Yeah.
- LFLex Fridman
... the best work of your life?
- RRRob Reid
Well, it would be very different if it's fiction versus non-fiction, and I've done both. I've written two works of non- two non-fiction books and two works of fiction. And two works of fiction being more recent, I'm gonna focus on that right now 'cause that's more toweringly on my mind.
- LFLex Fridman
(laughs)
- RRRob Reid
There, amongst novelists, again, this is an oversimplification, but there's kind of two schools of thought. Um, some people really like to fly by the seat of their pants, and some people really, really like to, to outline, to plot.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
You know, so there's plotters and pantsers, I guess is one way that people look at it.
- LFLex Fridman
(laughs)
- RRRob Reid
And, you know, as with most things, there is a great continuum in between, and I'm somewhere-
- LFLex Fridman
Yeah.
- RRRob Reid
... on that continuum, but I lean, I guess, a little bit m- a little bit more toward the plotter. And so when I do start a novel, I have a pretty strong point of view about how it's gonna end, and I have a very strong point of view about how it's gonna begin. And I do try to make an effort of making an outline that I know I'm gonna be extremely unfaithful to in the actual exe- execution of the story. But I'm trying to make an outline that gets us from here to there, and notion of subplots and beats and rhythm, and different characters and, and so forth. But then when I get into the process, that outline, particularly the center of it, ultimately, inevitably morphs a great deal, and I think if I were personally a rigorous outliner, I would not allow that to happen. I also would make a much more vigorous skeleton before I start. So I think people who are really in that plotting, outlining mode are people who write page-turners, people who write, you know, spy novels or, you know, supernatural adventures, where y- you really want a relentless pace of events, action, plot twists, conspiracy, et cetera, and that is really the bone, th- that's, that's really the, you know, the skeletal structure. So I think folks-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... who write that kind of book are really very much on the, the outlining side. And then I think people who write, um, what's often referred to as literary fiction, for lack of a better term, where it's more about, you know, sort of aura and ambiance and character development and experience and inner experience and inner journey and so forth, I think that group is more likely to fly by the seat of their pants. And I know people who start with a blank page and just see where it's gonna go. So I'm a little bit more on the plotting side. Um, now you asked what makes something, at least in the mind of the writer, as great as it can be. For me, it's an astonishingly high percentage of it is editing as opposed to the initial writing. For every hour that I spend writing new prose, you know, like new pages, new paragraphs, stuff that, you know, new bits of the book, I probably spend s- I mean, I wish I, I wish I kept a count. Like, I wish I had like one of those pieces of software that lawyers-
- LFLex Fridman
(laughs)
- RRRob Reid
... use to decide how much time I've been doing this or that. But I would say it's at least four or five hours, and maybe as many as 10, that I spend editing, and so it's relentless for me.
- LFLex Fridman
For each one-hour writing session?
- RRRob Reid
Mm-hmm, I'd say that.
- LFLex Fridman
Four. Wow.
- RRRob Reid
I'm a, I mean, I, I write because I edit, and I spend just relentlessly polishing and pruning, and sometimes on the micro level of just, like, does the, does the rhythm of the sentence feel right? D- do I need to carve a syllable or something so it can land? Like, as micro as that to as macro as, like, "Okay, I'm done, but the book is 750 pages long and it's way too bloated and I need to lop a third out of it." Problems on, you know, those two orders of magnitude and everything in between. Um, that is an enormous amount of my time, and I also, um, I also write music, write and record and produce music, and there the, the, the ratio is even higher, of every minute that I spend or my band spends laying down that original audio, there's a very high proportion of hours that go into just making it all hang together and sound just right. So I think that's true of a lot of creative processes, and I, I'm s- I know it's true of sculpture, um, I believe it's true of woodwork. My dad was an amateur woodworker and he spent a huge amount of time on sanding and polishing at the end, so I think a great deal of the sparkle comes from that part of the process, any creative process.
- LFLex Fridman
Can I ask about the psychological, the demon side of that picture? In the editing process, you're ultimately judging the initial piece of work-
- RRRob Reid
(laughs)
- LFLex Fridman
... and you're judging and judging and judging. Uh, how much of your time do you spend hating your work? How much time do you spend in gratitude, impressed, thankful for how good, uh, the work that you put together is?
- RRRob Reid
Um, I spend almost all the time in a place that's, um, intermediate between those, but leaning toward gratitude. I spend almost all the time in a state of optimism-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... that this thing that I have, I like, I like quite a bit, and I can make it better and better and better-
- LFLex Fridman
Mm-hmm.
- 18:54 – 26:40
Engineered viruses as a threat to human civilization
- LFLex Fridman
- RRRob Reid
Yeah.
- LFLex Fridman
But back to the, uh, destruction of human civilization. If humans destroy ourselves in the next 100 years, what will be the most likely source, the, the most likely reason that we destroy ourselves?
- RRRob Reid
Well, let's see. 100 years, it's hard for me to comfortably predict out that far, and it's something I give m- a lot more thought to, I think, than, you know, th- normal folks simply because I'm a science fiction writer.
- LFLex Fridman
Yeah.
- RRRob Reid
And, you know, I feel with the acceleration of technological progress, it's really hard to foresee out more than just a few decades. I mean, comparing today's world to that of 1921, where we are right now a century later, it would have been so unforeseeable. And I just don't know what's gonna happen, particularly with exponential technologies. I mean, our intuitions reliably defeat ourselves with exponential technologies like computing and synthetic biology, and, you know, how we might destroy ourselves in the honeyonear- 100-year timeframe might have everything to do with breakthroughs in nanotechnology 40 years from now and then how rapidly those breakthroughs accelerate. But in the nearer term that I'm comfortable predicting, let's say 30 years, I would say the most likely route to, to self-destruction would be synthetic biology.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And I always say that with a gigantic caveat, a very important one, that I find... And I'll, I'll abbreviate synthetic biology to synbio just to save us some syllables. I believe synbio offers us simply stunning promise that we would be fools to deny ourselves, so I'm not an anti-synbio person by any stretch. I mean, synbio has unbelievable odds of helping us beat cancer, helping us rescue the environment, helping us do things that we would currently find imponderable, so it's electrifying the field. But in the wrong hands, those hands either being incompetent or being malevolent, in the wrong hands, synthetic biology, to me, has a much, much greater odds, has much, much greater odds of leading to our self-destruction than something running amok with super AI, which I believe is a real possibility and one we need to be concerned about. But in the 30-year timeframe, I think it's a lesser one, uh, or nuclear weapons or anything else that I can think of.
- LFLex Fridman
Can you explain that a little bit further? So, your concern is on the manmade versus the natural side of the pandemic front here. So we humans engineering, uh, pathogens, engineering viruses-
- RRRob Reid
Right.
- LFLex Fridman
... is the, is the concern here?
- RRRob Reid
Yeah.
- LFLex Fridman
And maybe how do you see the possible trajectories happening here in terms of mo- is it malevolent or is it, um, accidents, oops, little mistakes, or unintended consequences of our particular actions that are ultimately lead to unexpected mistakes?
- RRRob Reid
Well, both of them are a danger, and I think the question of which is more likely has to do with two things. One, do we take a lot of methodical, affordable, foresighted steps that we are absolutely capable of taking right now to forestall the risk of a bad actor infecting us with something that could have annihilating impacts? And in the, the episode you referenced with Sam, uh, we talked a great deal about that. Um, so do we take those steps? And if we take those steps, I think the danger of malevolent rogue actors doing us in with synbio could plummet. But, you know, it's always a question of if, and we have a bad, bad, and very long track record of hitting the snooze bar after different natural pandemics have atta- have attacked us. So that's variable number one. Variable number two is how much experimentation and pathogen development do we as a society decide is acceptable in the realms of academia, government, or private industry? And if we decide as a society that it's perfectly okay for people with varying research agendas to create pathogens that, if released, could wipe out humanity, if we think that's fine, and if that kind of work starts happening in, you know, one lab, five labs, 50 labs, 500 labs, in one country, then 10 countries, then 70 countries or whatever, that risk of a booboo (laughs) starts rising astronomically. And this won't be a spoiler alert based on the way that I presented those two things, but I think it's unbelievably important to manage both of those risks. The easier one to manage, although it wouldn't be simple by any stretch because it would have to be something that all nations agree on, but the easiest way, the easier risk to manage is that of, "Hey, guys, let's not develop pathogens that if they escape from a lab could annihilate us." There's no line of research that justifies that, and in my view, I mean, that's the point of perspective we'd need to have. We'd have to collectively agree that there's no line of research that justifies that. The reason why I believe that would be a highly rational conclusion is even the highest level of biosafety lab in the world, Biosafety Lab level 4, and there are not a lot of BSL-4 labs in the world-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... there have, there have... Things can, can and have leaked out of BSL-4 labs. And some of the work that's been done with potentially annihilating pathogens, which we can talk about, is actually done at BSL-3. And so fundamentally any lab can leak. We have proven ourselves to be incapable of creating a lab that is utterly impervious to leaks. So why in the world would we create something where if, God forbid, it leaked, could annihilate us all? And by the way-... almost all of the measures that are taken in biosafety level anything labs are designed to prevent accidental leaks. What happens if you have a malevolent insider? And we could talk about the psychology and the motivations of what would make a malevolent insider who wants to release something an- annihilating in a bit. I'm sure that we will. But what if you have a malevolent insider? Virtually none of the standards that go into biosafety level one, two, three, and four are about preventing somebody hijacking the process. I mean, some of them are, but they're mainly designed against accidents, they're imperfect against accidents, and if this kind of work starts happening in lots and lots of labs, with every lab you add, the odds of there being a malevolent insider naturally increase arithmetically as the number of labs goes up. Now, on the front of somebody outside of a government academic, um, or scientific traditional government sci- academic scientific, um, environment creating something malevolent, again, there's protections that we can take both at the level of synbio architecture, the syn- hardening the entire synbio ecosystem against terrible things being made that we don't want to have out there by rogue actors, um, to early detection, to lots and lots of other things that we can do to dr- dramatically mitigate that risk.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And I think we do both of those things, decide that, no, we're not going to experimentally make annihilating pathogens in leaky labs, and B, yes, we are gonna take countermeasures that are cost- gonna cost a fraction of our annual defense budget to, to preclude their creation, then I think both ma- both, both risks get managed down. But if you take one set of precautions and not the other, then the, the, the thing that you have not taken precautions against immediately becomes the more likely
- 26:40 – 38:50
Gain-of-function research on viruses
- RRRob Reid
outcome.
- LFLex Fridman
So can we talk about this kind of research and what's actually done and w- what are the positives and negatives of it? So, uh, if we look at gain-of-function research and the kind of stuff that's happening level three and level four BSL labs, what's the whole idea here? Is it trying to engineer viruses to understand how they behave?
- RRRob Reid
Right.
- LFLex Fridman
You want to understand the dangerous ones.
- RRRob Reid
Yeah. So that, that would be the logic behind doing it. And so gain-of-function can mean a lot of different things. Um, viewed through a certain lens, gain-of-function research could be what you do when you create, you know, GMOs, when you create, you know, hardy strains of corn that are resistant to pesticides. I mean, you could view that as gain of function. So I'm gonna refer to gain of function in a relatively narrow sense, which is actually the sense that the term is usually used, which is in some way magnifying capabilities of microorganisms, uh, to make them more dangerous, whether it's more transmissible or more deadly. And in that line of research, I'll use an example from 2011, 'cause it's very illustrative and it's also very chilling. Back in 2011, two separate labs, independently of one another, I assume there was some kind of communication between them, but they were basically independent projects, one in Holland and one in Wisconsin, um, did gain-of-function research on something called H5N1 flu. H5N1 is, you know, something that, at least on a lethality basis, makes COVID look like a kitten. You know, COVID, according to the World Health Organization, has a case fatality rate somewhere between half a percent and 1%. H5N1 is closer to 60%, six zero. And so that's actually even slightly more lethal than Ebola. It's a very, very, very scary pathogen. The good news about H5N1 it is that it is barely, barely contagious, and I believe it is in no way contagious human-to-human. It requires, um, you know, very, very, very deep contact, uh, with birds, in most cases chickens. And so if you're a chicken farmer and you spend an enormous amount of time around them and perhaps you get into situations in which you get a break in your skin and you're interacting intensely with f- with fowl who, as it turns out, have H5N1, that's when the jump comes. Um, but it's not... there's no airborne transmission that we're aware of human-to-human. I mean, there, uh, not that we're- it just doesn't exist.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Um, I think the World Health Organization did a relentless survey of the number of H5N1, uh, cases. I think they do it every year. I saw one 10-year series where I think it was, like, 500 fatalities over the course of a decade, and that's a drop in the bucket, uh, kind of fun, fun fact. Um, I believe the typical lethality from lightning over 10 years is 70,000 deaths. So-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... we think getting struck by lightning, pretty low risk, H5N1, much, much lower than that. What happened in these experiments is the experimenters in both cases, um, set out to make H5N1 that would be contagious, so that could create airborne transmission. And so they basically passed it, I think in both cases they passed it through a large number of ferrets. And so this wasn't like CRISPR. There wasn't even any CRISPR back in those days. This was relatively straightforward, you know, selecting for a particular outcome, and after guiding the path and passing them through, a- again, I believe it was a series of ferrets, they did in fact come up with a version of H5N1 that is capable of airborne transmission. Now, they didn't unleash it into the world. They didn't, uh, inject it into humans to see what would happen. And so for those two reasons, we don't really know how contagious it might have been.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
But, you know, if it was as contagious as COVID, that could be a civilization-threatening pathogen. And why would you do it? Well, the people who did it were good guys. They were virologists. I believe their agenda as they explained it was, much as you said, "Let's figure out what a worst-case scenario might look like so we can understand it better." But, uh, my understanding is in both cases it was done in bio- BSL-3 labs, and so...... potential of leak, uh, significantly non-zero, hopefully way below 1%, but significantly non-zero, and when you look at the consequences of an escape in terms of human lives, destruction of a large portion of the economy, et cetera, and you do an expected value calculation on whatever fraction of 1% that was, you would come up with a staggering cost, staggering expected cost for this work. So it should never re- it should never have been carried out. Now, you might make an argument, if you said, if you believed that H5N1 in nature is on an inevitable path to airborne transmission, and it's only going to be a small number of years, A, and B, if it makes that transition, there is, you know, one set of changes to its metabolic pathways and, you know, its genomic code and so forth, one, that we have discovered. So, it is going to go from point A, which is where it is right now, to point B. We have reliably engineered point B. That is the destination. And we need to start fighting that right now because this is five years or less away.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Now, that'd be a very different world. That'd be like spotting an asteroid that's coming toward the Earth and it's five years off, and yes, you marshal everything you can to resist that. Well, there's two problems with that perspective. The first is, in however many thousands of generations that humans have been inhabiting this planet, there has never been a transmissible form (laughs) of H5N1, and influenza's been around for a very long time. So, th- th- there is no case for inevitability of this kind of a jump to airborne transmission. So w- w- we're not on a freight train to that outcome. And if there was inevitability around that, it's not like there's just one set of genetic code that would get there. There are just ... y- there's, there's all kinds of different mutations that could conceivably result in that kind of an outcome, unbelievable diversity of mutations. And so we're not actually creating something we're inevitably going to face, uh, but we are creating something, we are creating a very powerful and unbelievably negative card and injecting it into the deck that nature never should put into the deck. So, in that case, um, I just don't see any moral or scientific justification for that kind of work. And interestingly, there was, um, quite a bit of excitement and concern about this when the work came out. One of the teams was gonna publish their results in Science, the other in Nature. And there were a- a lot of editorials and a lot of scientists were saying, "This is crazy." And publication of those pla- papers did get suspended, and not long after that, there was a pause put on US government funding, NIH funding, on gain-of-function research. But both of those speed bumps were ultimately, uh, removed. Those papers did ultimately get published, and that pause on funding, you know, ceased long ago. And in fact, those two very projects, my understanding is, resumed their funding, got their government funding back, I don't know why a Dutch project's getting NIH funding, but whatever-
- LFLex Fridman
(laughs)
- RRRob Reid
... um, about a year and a half ago. So, as far as the US government and regulators are concerned, it's all sys- systems go for gain of function at this point, which I th- I find very troubling.
- LFLex Fridman
Now, I'm a little bit of an outsider from this field, but it has echoes of the same kind of problem I see in the AI world with autonomous weapons systems.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Nobody ... And, and my colleagues, my colleagues, friends, as far as I can tell, people in the AI community, are not really talking about autonomous weapons systems, as now US and China full steam ahead on the development of both.
- RRRob Reid
Right.
- LFLex Fridman
And that seems to be a similar kind of thing on gain of function. I've, uh, you know, have friends in the biology space, and they don't want to talk (laughs) about gain of function, uh, publicly. It ... And I don't ... That makes me very uncomfortable from an outsider perspective in terms of gain of function. It makes me, uh, very uncomfortable from the insider perspective on autonomous weapons systems. I'm not sure how to communicate exactly about autonomous weapons systems, and I certainly don't know how to communicate effectively about gain of function. What is the right path forward here? Should we cease all gain-of-function research? Is that, is that really the solution here?
- RRRob Reid
Well, again, I'm gonna use gain of function in the relatively narrow context of what we're discussing-
- LFLex Fridman
For ... Sorry, yeah. It's for viruses.
- RRRob Reid
... 'cause you, you could say almost, you know, anything that you do to make biology more effective is gain of function.
- LFLex Fridman
Yes.
- RRRob Reid
So within the narrow confines of what we're discussing, I, I think it would be easy enough for level-headed people in all of the countries, level-headed governmental people in all of the countries that realistically could support such a program to agree we don't want this to happen because all labs leak. I mean, and th- you know, an example that I, I use, I actually did use in the, the piece I did with Sam Harris as well, um, is the anthrax attacks in the United States in 2001. I mean, talk about an example of the least likely lab leaking into the least likely place. Uh, this was shortly after 9/11, for folks who don't remember it, and it was a very, very lethal strand of anthrax that, as it turned out, based on the fore- forensic genomic work that was done and so forth, absolutely leaked from a high security US army lab, probably the one at Fort Detrick in Maryland. Uh, it might have been another one, but who cares? It, it absolutely leaked from a high security US army lab. And where did it leak to, this highly dangerous substance that was kept under lock and key by a very security-minded organization? Well, it leaked to places including the Senate Majority Leader's office, Tom Daschle's office.
- LFLex Fridman
Yep.
- RRRob Reid
Think it was Senator Leahy's office, certain publications, including, bizarrely, The National Enquirer. But let's go to the Senate Majority Leader's office. It is hard to imagine a more security-minded country than the United States two weeks after the 9/11 attack. I mean, you ... It doesn't get more security minded than that. And it's also hard to imagine a more security capable organization than the United States military. We can joke all we want about inefficiencies in the military and, you know-... $24,000 wrenches and so forth, but pretty capable when it comes to that. Despite that level of focus and concern and competence, just a f- days after the 9/11 attacks, something comes from the inside of our military in- industrial complex and ends up, you know, in the office of someone, I believe a Senate majority leader, somewhere in the line of presidential succession. That tells us everything can leak. So again, think of a level-headed conversation between powerful leaders in a diversity of countries, thinking through, like I can imagine a very simple PowerPoint revealing, you know, just discussing briefly things like the anthrax leak, um, things like, uh, this, this foot and mouth disease out- outbreak that re- or out- leaking that came out of a BSL-4 level lab in the UK, several other things, talking about the utter virulence that could result from gain-of-function and say, "Folks, can we agree that this just shouldn't happen?" I mean, if we were able to agree on the Nuclear Non-Proliferation Treaty, which we were, Bio-Weapons Convention, which we did agree on, we, the world, for the most part, I, I believe agreement could be found there. But it's gonna take people in leadership of a couple of very powerful countries to get to consensus amongst them and then to decide we're gonna get everybody together and browbeat them into banning the stuff. Now, that doesn't make it entirely impossible that somebody might do this, but in well-regulated, you know, carefully watched over fiduciary environments, like federally funded academic research, anything going on in the government itself, you know, things going on in, you know, companies that have investors who don't want to go to jail for the rest of their lives, um, I think that would have a major, major dampening impact on it.
- 38:50 – 46:10
Did COVID leak from a lab?
- RRRob Reid
- LFLex Fridman
But there is a, a particular possible catalyst in this time we live in, which is, uh, for really kind of raising the question of gain-of-function research for the application of virus, making viruses more dangerous, is the question of whether COVID leaked from a lab. The s- sort of not even answering that question, but even asking that question is a very, it seems like a very important question to ask to, uh, catalyze the conversation about whether we should be doing gain-of-function research. I mean, from a high level, uh, why do you think people, y- even colleagues of mine, are not comfortable asking that question? And two, do you think that the answer could be that it did leak from a lab?
- RRRob Reid
I, I think the mere possibility that it did leak from a lab is evidence enough, again, for the hypothetical rational national leaders watching this simple PowerPoint. If you could put the possibility at 1% and you look at the unbelievable destructive power that COVID had, that should be an overwhelmingly powerful argument for excluding it. Now, as to whether or not that was a leak, some very, very level... I, I don't, I don't know enough about all of the factors and the Bayesian analysis and so forth that has gone into people making the pro argument of that. So, I don't pretend to be an expert on that and I, I don't have r- point of view. I, I just don't know. But what I, what we can say is it is entirely possible for a couple of reasons. One is that there is a BSL-4 lab in Wuhan, the Wuhan Institute of Virology. I believe it's the, the only BSL-4 in China. I could be wrong about that. But it definitely had a history that alarmed very sophisticated, uh, US diplomats and others who were in contact with the lab and were aware of what it was doing, uh, long before COVID, uh, COVID, um, hit the world. And so, there are diplomatic cables that have been declassified. I th- I believe one sophisticated scientist or other observer said that WIV is a ticking time bomb, and I believe it's also been pretty reasonably established that coronaviruses were a topic of great interest at WIV. Uh, SARS obviously came out of China, and that's a, that's a coronavirus, so it would make an enormous amount of sense for it to be studied there. Um, and there is so much opacity about what happened in the early days and weeks after the outbreak that's basically been imposed by the Chinese government that w- we just don't know. So, it, it feels like a substantially or greater than 1% possibility to me looking at it from the outside, and that's something that one could imagine. Now, we're going to the realm of thought experiment, not me decreeing this is what happened, but, you know, if they're studying coronavirus at the Wuhan Institute of Virology, um, and there is this precedent of gain-of-function research that's been done on something that is remarkably uncontagious to humans, whereas we know coronavirus is contagious to humans. I could definit- and, and there has been global consensus, you know. Certainly was the case, you know, two or three years ago when this work might have started, there seems to be this global consensus that gain of function is fine. The US paused funding for a little while, but paused funding. They never said, "Private actors couldn't do it."
- LFLex Fridman
Mm-hmm.
- RRRob Reid
It was just a pause of NIH funding, and then that pause was lifted. So, uh, again, none of this is irrational. You could certainly see the folks at WIV saying, "Hmm, gain of function, interesting vector. Coronavirus, unlike H5N1, very contagious. Uh, we are in a, uh, uh, a nation that has had terrible run-ins with coronavirus. Why don't we do a little gain of function on this?" And then, like all labs at all levels, one could imagine this lab leaking. So, it's not an impossibility and very, very level-headed people have said that, you know, who've looked at it much more deeply, do believe in that outcome. Um-
- LFLex Fridman
Why is it such, such a threat to power, the idea that it leaked from a lab? Why is it so threatening? I don't maybe understand this point exactly.
- RRRob Reid
Mm.
- LFLex Fridman
Like...I- is it just that as governments, and especially the Chinese government, is really afraid of admitting mistakes that everybody makes? So this is a horrible mis-
- RRRob Reid
Yeah.
- LFLex Fridman
... like, uh, Chernobyl is a good example. I come from the Soviet Union. I mean, well, major mistakes were made in Chernobyl. I would argue for a lab leak to happen, the, the, the scale of the mistake is much smaller.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Right? There, (laughs) the, the depth and the breadth of, uh, uh, rot that, in bureaucracy, that led to Chernobyl is much bigger than anything that could lead to a, a lab leak. 'Cause it's, it could literally just be, I mean, I'm sure there's security, very careful security procedures even in level three labs, but it, uh, I- I- I imagine, maybe you can correct me-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... it's, all it takes is the incompetence of a small number of individuals.
- RRRob Reid
Or even one, yeah.
- LFLex Fridman
Uh, one individual on a particular, uh, uh, couple weeks, three weeks period, as opposed to a multi-year bureaucratic failure of the entire government.
- RRRob Reid
Right. Well, certainly the magnitude of mistakes and compounding mistakes that went into Chernobyl was far, far, far greater.
- LFLex Fridman
Yeah.
- RRRob Reid
But the consequence of COVID outweighs that, the consequence-
- LFLex Fridman
Gotcha.
- RRRob Reid
... of Chernobyl to a tremendous degree. And, you know, I think that the, that particularly, um, authoritarian governments are unbelievably, uh, reluctant to admit to any fallibility whatsoever. I mean, there's a l- long, long history of that across dozens and dozens of authoritarian governments. And to be transparent, again, this is in the hypothetical world in which this was a leak, which again, I don't have, I don't personally have enough sophistication to have an opinion on the re- on the likelihood. But in the hypothetical world in which it was a leak, the global reaction and the amount of global animus and the amount of, you know, the decline in global respect that would happen toward China, because every country suffered massively from this, unbelievable damages in terms of human lives and economic activity disrupted, um, the world would in some way present China with that bill.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And when you take on top of that the natural disinclination for any authoritarian government to admit any fallibility and tolerate the possibility of any fallibility whatsoever, um, and you look at the relative opacity, even though they let a World H- Health Organization group in, you know, a couple months ago to run around, they didn't give that WHO group anywhere near the level of access it would be necessary to definitively say, "X happened," versus, "Y." The level of opacity that surrounds those opening weeks and months of COVID in China, w- we just don't know.
- 46:10 – 53:59
Virus detection
- LFLex Fridman
If you were to kind of look back at 2020 and maybe broadening it out to future pandemics-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... that could be much more dangerous, what kind of response, how do we fail in a response, and how could we do better?
- RRRob Reid
Mm-hmm.
- LFLex Fridman
So the gain-of-function research is discussing we sh- you know, the, the question of we should not be creating viruses that are both exceptionally contagious and exceptionally deadly to humans. But if it does happen, perhaps through natural evolution, ma- natural mutation-
- RRRob Reid
Yeah.
- LFLex Fridman
... h- is there interesting technological responses on the testing side, on the vaccine development side, on the collection of data, or on the basic sort of policy response side, or the sociological, the psychological side?
- RRRob Reid
Yeah, there's all kinds of things. And m- most of what I've thought about and written about and, again, discussed in that long bit with, with Sam is dual use. So most of the countermeasures that I've been thinking about and advocating for would be every bit as effective against a zoonotic disease, a natural pandemic-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... uh, of some sort, as an artificial one. The, the risk of an artificial one, even the near-term risk of an artificial one, ups the urgency around these measures immensely, but, but most of them would be broadly applicable. And so I think the first thing that we really want to do on a global scale is have a far, far, far more robust and globally transparent system of detection, and that can happen on a number of levels. The most obvious one is, you know, just in the blood of people who come into clinics exhibiting signs of illness.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And there, we, we are certainly at a point now with, we're at, with relatively minimal investment, we could develop in-clinic diagnostics that would be unbelievably effective at pinpointing what's going on in almost any disease when somebody walks into a doctor's office or a clinic. And better than that, um, this is a little bit further off, further off, but it wouldn't cost tens of billions in research dollars, it would be, you know, a relatively modest and affordable budget in relation to the threat, at-home diagnostics that can really, really pinpoint, you know, okay, particularly with respiratory infections, because that is generally almost universally the mechanism of transmission for any serious pandemic. So somebody has a respiratory infection, is it one of the, you know, significantly large handful of rhinoviruses, coronaviruses, and other things that, that cause common cold, uh, or is it influenza? If it's influenza, is it influenza A versus B? Um, or is it, you know, a small handful of other more exotic, but nonetheless sort of common respiratory infections that are out there? Developing a diagnostic panel to pinpoint all of that stuff, that's something that's well within our capabilities. That's much less will list- uh, lift than creating mRNA vaccines, which obviously we proved capable of when we put our minds to it. So do that on a global basis. And I don't think that's irrational, because the best prototype be-... for this than I'm aware of, isn't currently rolling out in Atherton, California or Fairfield County, Connecticut, or some other wealthy place. The best prototype that I'm aware of this is rolling out right now in Nigeria, and it's a project that came out of the Broad Institute, um, which is, as- as I'm sure you know but, uh, some listeners may not, is kind of like an academic joint venture between Harvard and MIT. The program is called Sentinel and their objective is, and th- their plan, and it's a very well-conceived plan, a methodical plan, is to do just that. In areas of Nigeria that are particularly vulnerable to zoonotic, uh, diseases making the jump from animals to humans. But also there's just an unbelievable ha- public health benefit from that. And it's sort of a three-tier system where clinicians in the field could very rapidly determine, do you have one of the infections of acute interest here, either because it's very common in this region so we want to diagnose as many as- things as we can at the frontline, or because it's uncommon but unbelievably threatening, like Ebola. So, frontline worker can make that determination very, very rapidly. If it comes up as a, "We don't know," they bump it up to a level that's more like at a fully configured doctor's office or local hospital. And if it's still at a, "We don't know," it gets bumped up to a national level. And that... and- and it gets bumped very, very rapidly. So, if this can be done in Nigeria, and it seems that it can be, uh, there shouldn't be any inhibition for it to happen in most other places, and it should be affordable from a budgetary standpoint. And based on Sentinel's budget and adjusting things for things like, you know, very different cost of living, larger population, et cetera, I did a back of the envelope calculation that doing something like Sentinel in the US would be in the low billions of dollars. And, you know, wealthy countries, middle income countries can afford such a thing. Lower-incan companies... in- income countries should certainly be helped with that. But start with that level of detection and then layer on top of that other interesting things like, you know, monitoring search engine traffic, search engine queries for evidence that strange clusters of symptoms are starting to rise in different places. There's been a lot of work done with that, most of it kind of ac- academic and experimental, but some of it has been powerful enough to suggest that this could be a very powerful early warning system. There's a guy named Bill Lampos at University College London, um, who basically did a very rigorous, um, analysis that showed that symptom searches reliably predicted COVID outbreaks in the early days of the pandemic in given countries by as much as 16 days before the evidence started to accrue at a public health level. 16 days of forewarning can be monumentally important in the early days of an outbreak, and this is, you know, a very, very talented but nonetheless very resource constrained academic project. Imagine if that was something that was done with a NORAD-like budget.
- LFLex Fridman
Yeah.
- RRRob Reid
Yeah. So, I mean, starting with detection, that's something we could do radically, radically better.
- LFLex Fridman
So aggregating multiple data sources in order to create something... I mean, this is really exciting to me, the possibility that I've heard inklings of-
- RRRob Reid
Mm.
- LFLex Fridman
... of creating almost like a weather map of pathogens. Like, um, basically aggregating all of these data sources, scaling many orders of magnitude up at home testing and all kinds of testing that doesn't just try to test for the particular pathogen of worry now but everything-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... like a full spectrum of things that could be dangerous to the human body and thereby be able to create these maps, like, that are dynamically updated on a hourly basis of th- of how viruses travel throughout the world, and so you can respond... Like you can then integrate, just like you do when you check your weather map-
- RRRob Reid
Mm.
- LFLex Fridman
... and it's raining or not. Of course, not perfect, but it's very good predictor of whether it's gonna rain or n- not, uh, and use that to then make decisions about your own life. Ultimately give the power and information to individuals to respond. And if it's a super dangerous, like if it's acid rain (laughs) versus regular rain, you might want to really stay inside as opposed to risking it. And that, um, e- just like you said, if... I think it's not very expensive relative to all other things that we do in this world, but it does require bold leadership.
- 53:59 – 1:01:43
Failure of institutions
- LFLex Fridman
And there's another dark thing which really has bothered me about 2020 which it requires, is it requires trust in, um, institutions to carry out these kinds of programs, and it requires trust in science and engineers and, um, sort of centralized organizations that would operate at scale here. And much of that trust has been, um, at least in the United States, diminished, it feels like. I'm not exactly sure where to place the blame but I do place quite a bit of the blame into the scientific community, and again, my fellow colleagues, in speaking down to people at times, speaking from authority. It sounded like it dismissed the basic human experience or the- the basic common humanity of people in a way that, like, it almost sounded like there's an agenda that's hidden behind the words the scientist spoke, like they're trying to, in a self-preserving way, control the population or something like that. I don't think any of that is true from the majority of the scientific community but it sounded that way, and so the trust s- began to diminish and I'm not sure how to fix that except to be more authentic, be more real, acknowledge the uncertainties under which we operate, acknowledge the mistakes we've, uh, that, uh, scientists make, that institutions make. The leak from the lab was a perfect example where we have imperfect systems that make all the progress we see in the world, and that... being honest about that imperfection, I think, is essential for forming trust. But I don't know what to make of it. It's been, uh...... been deeply disappointing because I do think, just like you mentioned, the solutions require people to trust the institutions with their data. (laughs)
- RRRob Reid
Yeah, and I, I think part of the problem is, it seems to me as an outsider, that there was a bizarre unwillingness on the part of the CDC and other institutions to, uh, admit to, to frame, and to contextualize uncertainty. Maybe they had a patronizing idea that these people need to be told, and when they're told, they need to be told with authority and a level of definitiveness and cert- certitude that doesn't actually exist.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And so, when they whipsaw on recommendations like w- what you should do about masks, you know, when the CDC is kind of at the very beginning of the pandemic saying, "Masks don't do anything. Don't wear them," when the real driver for that was, "We don't want these clowns going out and depleting Amazon of masks because they may be needed in medical settings, and we just don't know yet," I think a message that actually respected people and said, "This is why we're asking you not to do masks yet w- and there's more to be seen," would be less whipsawing and would bring people, like they feel more like they're part of the conversation and they're being treated like adults than saying one day, definitively, "Masks suck," and then X days later saying, "Nope, damn it, wear masks." And so I think framing things in terms of the probabilities, which most people are easy to parse. I mean, a- a more recent example, which I just thought was batty, was suspending the Johnson & Johnson vaccine for a, you know, a l- very low single digit number of days in the United States based on the fact that, I believe, there had been, uh, seven-ish clotting incidents, um, in roughly 7 million people who had had the, the vaccine administered, I believe one of which resulted in a fatality. And there was definitely suggestive data that indicated that there was a relationship, this wasn't just coincidental, because I think all of the clotting incidents happened in women as opposed to men and kind of clustered in a certain age group. But does that call for shutting off the vaccine or does it call for leveling with the American public in saying, "We've had one fatality out of 7 million. This is, let's just assume, substantially less than the, the likelihood of getting struck by lightning. Um, based on that information, you know, and we're gonna keep you posted 'cause you can trust us to keep you posted, based on that information, please decide whether you're comfortable with a Johnson & Johnson vaccine." That would have been one response, and I think people would have been able to parse those simple bits of data and make their own judgment. By turning it off, all of a sudden there's this dramatic signal to people who don't read all 900 words in the New York Times piece that explains why it's being turned off, but just see the headline, which is a majority of people. There's a sudden like, "Oh, my God. Yikes."
- LFLex Fridman
Mm-hmm.
- RRRob Reid
"Vaccine being shut off." And then all the people who sat on the fence or are sitting on the fence about whether or not they trust vaccines, that is gonna push an incalculable number of people. That's gonna be the last straw for we don't know how many hundreds of thousands or more likely millions of people to say, "Okay, tipping point here. I don't trust these vaccines." By pausing that for, whatever it was, 10 or 12 days, and then flipping the switch as everybody who knew much about the situation knew was inevitable. By switching, flipping the on switch 12 days later, you're conveying certitude J&J bad to certitude J&J good in a period of just a few days and people just feel whipsawed and they're-
- LFLex Fridman
Yeah.
- RRRob Reid
... not part of the a- the analysis.
- LFLex Fridman
But it's not just the, the whipsawing, and I think about this quite a bit.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
I don't think I have good answers. It's something about the way the communication actually happens. Just, I don't know what it is about Anthony Fauci, for example, but I don't trust him.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
And I think that has to do, I mean, he's, he did, he's, uh, he has an incredible background. I'm sure he's a brilliant scientist and researcher. I'm sure he's also a great, uh, like inside the room policymaker and deliberator and so on. But, you know, uh, what makes a great leader is something about that thing that you can't quite describe, but being a communicator that you know you can trust a- that there's an authenticity that's required. And I'm not sure, maybe I'm being a bit too judgmental, but I'm a huge fan of a lot of lea- great leaders throughout history. They've, they've communicated exceptionally well in the way that Fauci does not, and I think about that. I, I think about what is effective science communication. So, you know, great leaders throughout history did not necessarily need to be great science communicators.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Their leadership was in, in, in the other domains. But when you're fighting the virus, you also have to be a great science communicator. You have to be able to communicate uncertainties. You have to be able to communicate something like a vaccine that you, you're allowing inside your body into the messiness and to the complexity of the biology system, that if we're being honest, is so complex we'll never be able to really understand. We have, uh, we can only desperately hope that science can give us sort of a, a high likelihood that there's no short-term negative consequences and a kind of intuition about long-term negative consequences and doing our best in this battle against trillions of things that are trying to kill us. I mean, being a, uh, being a, an effective communicator in that space is very difficult, but, uh, I think about what it takes because I think there should be more science communicators that are effective at that kind of thing.
- 1:01:43 – 1:06:02
Using AI to engineer viruses
- LFLex Fridman
Let me ask you about something that's sort of more in the AI space that-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
I, I think about, that kind of goes along this thread that you're, that you've spoken about, about democratizing the technology that could destroy human (laughs) civilization-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... is, uh, from amazing work from DeepMind, AlphaFold2, which achieved, uh, incredible performance on the protein folding problem, single protein fold, folding problem. Do you think about the use of AI in the synbio space of, uh, I think the, the gain-of-function in the virus space research that you referred to, I think is natural mutations and sort of aggressively mutating the virus until you get one that, like, uh, that has ... that's both contagious and, um, deadly. But what about then using AI to, through simulation be able to compute deadly viruses or any kind of biological systems? Is this something you're worried about? Or, again, is this something you're more excited about?
- RRRob Reid
I, I think computational biology is unbelievably exciting and promising field, and I think when you're doing things in silico as opposed to in vivo, um, you know, the, the dangers plummet. You, you don't have a critter that can leak from a leaky lab.
- LFLex Fridman
Yes.
- RRRob Reid
So I don't see any problem with that except, um, I do worry about the data security dimension of it because if you were doing really, really interesting in silico gain-of-function research and you hit upon, you know, through a level of sophistication we don't currently have but, you know, synthetic biology is an exponential technology, so capabilities that are utterly out of reach today will be attainable in five or six years. Um, I think if you conjured up worst-case genomes of viruses that don't exist in vivo anywhere, they're just in, they're just in the computer space, but, like, "Hey, guys, this is the genetic sequence that would end the world," let's say.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Um, then you have to worry about the utter hackability of every computer network we can imagine. I mean, data leaks, um, from the least likely places on the grandest possible scales have happened and continue to happen and will prob- probably always continue to happen. And so that would be the danger of doing the work in silico. Um, if you end up with a list of, like, "Well, these are things we never want to see," and that list leaks and after the passage of some time, certainly couldn't be done today, but after the passage of some time, um, lots and lots of people in academic labs going all the way down to the high school level are in a position to, you know, to make it overly simplistic, hit print on a genome and have the virus bearing that genome pop out on the other end, then you got something to worry about. But, in general, computational biology, I think, is incredibly important, particularly because the crushing majority of work that people are doing with the protein folding problem and other things are about creating therapeutics, about creating things that will help us, you know, live better, live longer, thrive, be b- more well, and so forth. And the protein folding problem is a monstrous computational challenge that we seem to make just the most glacial project on, I'm sorry, progress on for years and years. But I think there's, like, a, there's a biannual competition, I think, um, for, for, at which people tackle the protein folding, uh, problem and, um, DeepMind's entrant, uh, both two years ago, like, in 2018 and 2020 ruled the field.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And so, you know, pro- protein folding is an unbelievably important thing if you want to start thinking about th- therapeutics because, you know, it's the folding of the protein that tells us where the, where the channels and the receptors and everything else are on that protein and it's from that precise model, if we can get to a precise model, that you can start barraging it again in silico with, you know, thousands, tens of thousands, millions of potential therapeutics and see what resolves the problems, the shortcomings that a, you know, a ba- a misshapen, uh, pro- protein, for instance, in somebody with cyst- cystic fibrosis, how might we treat that? So, I see nothing but good in that.
- LFLex Fridman
Well,
- 1:06:02 – 1:15:21
Evil and competence
- LFLex Fridman
let me ask you about fear and hope-
- RRRob Reid
Mm.
- LFLex Fridman
... in this world. (laughs)
- RRRob Reid
(laughs)
- LFLex Fridman
I tend to believe that, um, that, uh, in terms of competence and malevolence-
- RRRob Reid
Mm.
- LFLex Fridman
... that people who are, and maybe it's in my interactions, I tend to see that, first of all, I, I believe that most people are good and want to do good and are just better at doing good and more in- uh, inclined to do good on this world. And then more than that, people who are malevolent are usually incompetent at, uh, building technology. So, like, I, I've seen this in my li- life that people who are exceptionally good at stuff, no matter what the stuff is, tend to maybe they discover joy in life in a way that gives them fulfillment and thereby does not result in them wanting to destroy the world. (laughs) So, like, the better you are at stuff, whether that's building nuclear weapons or plumbing, doesn't matter, they're both, the less likely you are to destroy the world. So, in that sense, with many technologies, AI especially, I always think that, uh, the, the, the malevolent will be far outnumber by the ultra-competent, and in that sense, the defenses will always be stronger than the offense in terms of the people trying to destroy the world. Now, there is a few spaces where that mi- that might not be the case, and that's an interesting conversation where this, this one person who's not very competent can destroy the whole world. Perhaps synbio is one such space because of the, uh, exponential effects of the technology. I tend to believe AI is, is not one of those such spaces, but-Do you- do you share this kind of view that, uh, the ultra-competent are usually also the good?
- RRRob Reid
Yeah, absolutely. Uh, I absolutely share that and that gives me a great deal of optimism that we will be able to short-circuit the threat that malevolence and bio could pose to us. But we need to start creating those defensive systems-
- LFLex Fridman
Yeah.
- RRRob Reid
... or defensive layers, one of which we talked about, far, far, far better surveillance, in order to prevail. So, the good guys will almost inevitably outsmart and definitely outnumber the bad guys in most sort of smackdowns that we can imagine. But the good guys aren't going to be able to exert their advantages unless they have the imagination necessary to think about what the worst possible thing can be done by somebody whose own psychology is completely alien to their own. So, that's a tricky, tricky thing to solve for. Now, in terms of whether the asymmetric power that a bad guy might have in the face of the overwhelming numerical advantage and competence advantage that the good guys have, you know, unfortunately I look at something like mass shootings as an example. You know, I'm sure the guy who- who was responsible for the Vegas shooting or the Orlando shooting or any other shooting that we can imagine didn't know a whole lot about ballistics. And the number of, you know, good guy citizens in the United States with guns compared to bad guy citizens I'm sure is a crushingly, overwhelmingly high ratio in favor of the good guys. But that doesn't make it possible for us to stop mass shootings. Um, an example is Fort Hood, um, 45,000 trained soldiers (laughs) on that base yet there have been two mass shootings there. And so there is an asymmetry when you have powerful and lethal technology that gets so democratized and so proliferated in tools that are very, very easy to use, even by a knucklehead. When those tools get really easy to use by a knucklehead and they're really widespread, it becomes very, very hard to defend against all instance in- instances of usage. Now, the "good news" quote unquote about mass shootings, if there is any, and there is some, is even the most brutal and carefully planning and well-armed mass shooter can only take so many victims. And same is true as, th- there's been four instances that I'm aware of of commercial pilots committing suicide by downing their planes and taking all their passengers with them. These weren't Boeing engineers, you know, but, like, an army of Boeing engineers ultimately were not capable of preventing that. But even in their case... And I'm actually not counting 9/11 in that. 9/11's a- a different category in my mind. These are- these are just personally suicidal pilots. In- in those cases, they only have a planeload of people that they're able to take with them. If we imagine a highly plausible and imaginable future in which some bio tools that could be- that are amoral, that could be used for good or for ill start embodying unbelievable sophistication and genius in the tool, in the easier and easier and easier to make tool, all those thousands, tens of thousands, hundreds of thousands of scientist years start getting embodied in something that, you know, may be as simple as hitting a print button.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Um, then that good guy technology can be hijacked by a bad person-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... and used in a very asymmetric way.
- LFLex Fridman
See, I think what happens though as y- is g- 'cause you go to- to the high school student from the current, like, very specific set of, uh, labs, they're able to do it.
- RRRob Reid
Yeah.
- LFLex Fridman
As we get... As it becomes more and more democratized, as it becomes easier and easier to do this kind of large scale damage with a- with an engineered virus, the more and more there will be engineering of defenses against these systems. It's some of the things we talked about in terms of testing, in terms of collection of data, but also in terms of, like, uh, at- at scale contact tracing or also engineering of vaccines.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Like, in a matter of, like, days, maybe hours, maybe minutes. So, like, I- I just- I feel like the defenses... This is what human species seems to do, is like we s- hit- keep hitting the snooze button until there's, like, a- like, a storm on the horizon heading towards us then we start to quickly build up, uh, the defenses or the response that's proportional to the scale of the storm. Uh, of course, again, certain kinds of exponential threats require us to build up the defenses, uh, way earlier than we usually do and that's, I guess, the question. But I ultimately am hopeful that the natural process of hitting the snooze button until the deadline is right in front of us will work out for quite a long time for us humans 'cause-
- RRRob Reid
And I fully agree. I mean, that's why I'm fundamentally... It may not sound like it thus far, but I'm fundamentally very, very optimistic about-
- LFLex Fridman
Yes.
- RRRob Reid
... our ability to short-circuit this threat because there is, again, I'll stress, um, the technological feasibility and the profound affordability of a relatively simple set of steps that we can take to preclude it, but we do have to take those steps.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
And so, you know, what I'm hoping to do and trying to do is inject a notion of what those steps are, you know, into the public conversation and do my small part to up the odds that that actually ends up happening. Um, y- you know, it's... The- the- the danger with this one is it is exponential and I think that our minds are f- fundamentally struggle to understand exponential math. It's just not something we're wired for. Our ancestors didn't confront exponential processes when they were growing up on the savanna, so it's not something that's intuitive to us and our intuitions are reliably defeated when exponential processes c- come along. So that- that's issue number one. And issue number two with something like this is, you know, it kind of only takes one.You know, that ball only has to go into the net once and we're doomed, which is not the case w- with mass shooters. It's not the case with, you know, commercial pilots run amok. It's not the case with really any threat that I can think of, with the exception of nuclear war, that has the, you know, one bad outcome and game over. And that- that means that we need to be unbelievably serious about these defenses, and we need to do things that might on the surface seem like a tremendous overreaction so that we can be prepared to nip anything that comes along in the bud. But I, like you, believe that's eminently doable. Um, I, like you, believe that the good guys outnumber the bad guys in this particular one to a degree that probably has no precedent in history. I mean, even the worst, worst people, I'm sure, in ISIS, even Osama bin Laden, even any bad guy you could imagine in history would be revolted by the idea of exterminating all of humanity. I mean, you know, th- that's just, that's a low bar.
- LFLex Fridman
(laughs)
- RRRob Reid
Um, and so the good guys completely outnumber the bad guys when it comes to this, but th- the asymmetry and the fact that one catastrophic...
- LFLex Fridman
Right.
- RRRob Reid
... error could lead to unbelievably consequential things is what worries me here. But I too am very optimistic.
- 1:15:21 – 1:19:14
Where are the aliens?
- RRRob Reid
- LFLex Fridman
The thing that I sometimes worry about is, uh, the fact that we haven't seen overwhelming evidence of alien civilizations out there.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Makes me think, um, well, there's a lot of explanations, but one of them that worries me is that, uh, whenever they get smart, they just destroy themselves.
- RRRob Reid
Oh, yeah. I mean, that was the most fascinating, is the most fascinating and chilling number, or variable, in the Drake equation, is L. At the end of-
- LFLex Fridman
Yeah.
- RRRob Reid
... at the end of it, you look out you see, you know, one to 400 billion stars in the Milky Way galaxy and we now know because of Kepler that an astonishingly high percentage of them probably have habitable planets, and, you know, so all the things that were unknowns when the Drake equation was originally written, like, you know, how many stars have planets? Actually back then, in the 1960s when the Drake equation came along, the consensus amongst astronomers was that it would be a small minority of solar systems that had planets-
- LFLex Fridman
Right.
- RRRob Reid
... with, with stars, but now we know it's substantially all of them. How many of those stars have habitable, have planets in the habitable zone? It's kind of looking like 20%. Like, oh my god. And so L, which is how long does a civilization once it reaches technological competence continues to last-
- LFLex Fridman
Yeah.
- RRRob Reid
... that's the doozy.
- LFLex Fridman
(laughs)
- RRRob Reid
And, and, and, and you're right.
- LFLex Fridman
Yeah.
- RRRob Reid
Uh, it's, it, it's all too plausible to think that when a civilization reaches a level of sophistication that's probably just a decade or three in our future, the odds of it self-destructing just start mounting astronomically, no pun intended.
- LFLex Fridman
(laughs) My, my hope is that, that, uh, actually there is a lot of alien civilizations out there and what they figure out in order to avoid the self-destruction, they need to turn off the thing that was useful, that used to be a feature, now became a bug which is, uh, the desire to colonize.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
To conquer more land, to... So they, like, there's probably ultra-intelligent alien civilizations out there that are just, like, chilling, like, on the beach with the, with the whatever your favorite alcohol bever- beverage is. But, like, without sort of trying to conquer everything. Just chilling out and maybe e- exploring in the, in the realm of knowledge, but almost like appreciating existence for its own sake versus, uh, life as a progression of conquering of other life.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
Like this kind of predator-prey formulation that resulted in, uh, us humans perhaps is, uh, something we have to shed in order to survive. I don't know.
- RRRob Reid
Yeah, that, that is, um, a very plausible solution to Fermi's paradox, and it's one that makes sense. You know, when we look at our own lives and our own arc of traje- of technological, you know, trajectory, it's very, very easy to imagine that in an intermediate future world of, you know, flawless VR or flawless, you know, whatever kind of simulation that we want to inhabit, it will just simply cease to be worthwhile-
- LFLex Fridman
Yeah.
- RRRob Reid
... to go out and, and y- expand our ge- our interstellar territory. And, but if we were going out and conquering interstellar territory, it wouldn't necessarily have to be predator or prey. I could imagine, um, a benign but sophisticated intelligence saying, "Well, we're going to go to places, we're going to go to places that we can terraform."
- LFLex Fridman
Yeah.
- RRRob Reid
Use a different word than terra-, obviously, but we can turn into habitable for our particular physiology.
- LFLex Fridman
Yeah.
- RRRob Reid
Uh, so long as they don't house, you know, intelligent sentient creatures that would suffer from our invasion.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
Um, but it is easy to see a sophisticated intelligent species evolving to the point where interstellar travel, with its incalculable expense and physical hurdles, just isn't worth it compared to what could be done, you know, where one already is.
- 1:19:14 – 1:28:43
Backing up human consciousness by colonizing space
- RRRob Reid
- LFLex Fridman
So, you talked about diagnostics at scale as a possible solution to, uh, future pandemics. Um, what about another possible solution which is, uh, kind of creating a backup copy, you know? I'm actually now, um, uh, putting together and asked for a backup for myself for the first time, taking backup of data seriously.
- RRRob Reid
Mm-hmm.
- LFLex Fridman
But if we were to take the, uh, the backup of human consciousness seriously, and, uh, try to expand throughout the, uh, solar system and colonize other planets, do you think that's an interesting, uh...... solution, one of many, uh, for, uh, protecting human civilization from self-destruction, sort of humans becoming a multi-planetary species?
- RRRob Reid
Oh, absolutely. I mean, I find it electrifying, first of all, so I've got a little bit of a personal bias. When I was a kid, I thought there was nothing cooler than rockets. I thought there was nothing-
- LFLex Fridman
(laughs)
- RRRob Reid
... cooler than NASA. I thought there was nothing cooler than people walking on the moon. And as I grew up, um, I thought there was nothing more tragic than the fact that we went from walking on the moon to, at best, getting to something like sub-orbital altitude. And just, I found that more and more depressing with the passage of decades, at just the colossal expense of, you know, manned space travel, and the fact that it seemed that we were unlikely to ever get back to the moon, let alone Mars. So, I have a boundless app- appreciation for Elon Musk for many reasons, but the fact that he has put Mars on the credible agenda is one of the things that I appreciate immensely. So, there's just this sort of space nerd in me that just says, "God, that's cool." But on a more practical level, we were talking about, you know, uh, potentially inhabiting planets that aren't our own, and we're thinking about a benign civilization that would do that in, in planetary circumstances where we're not causing other conscious systems to suffer. I mean, Mars is a place that's very promising. There may be microbial life there, and I hope there is, and if we found it, I think it would be electrifying. But I think ultimately, the moral judgment would be made that, you know, the continued thriving of that microbial life is of less concern than creating a habitable planet to humans, which would be a project on the many thousands of years scale. But I don't think that that would be a greatly immoral act. And if that happened, and if Mars became, you know, home to a self-sustaining group of humans that could survive a catastrophic mistake here on Earth, then yeah, the fact that we have a backup colony is great, and if we could make more... I'm sorry, not backup colony, backup copy is great. And if we can make more and more such backup cop- copies throughout the solar system by hollowing out asteroids and whatever else it is-
- LFLex Fridman
Yeah.
- RRRob Reid
... maybe even Venus, we could get rid of three-quarters of its atmosphere and, you know, turn it into a tropical paradise. Um, I think all of that is wonderful. Now, whether we can make the leap from that to interstellar trans- transportation with the incredible distances that are involved, um, I think that's an open question. But I think if we ever do that, it would be more like the Pacific Ocean's, uh, channel of human expansion than the Atlantic Ocean's. And so, what I mean by that is, uh, when we think about European society transmitting itself across the Atlantic-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... it's these big, ambitious, crazy expensive one-shot expeditions, like Columbus'-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... to make it across this enormous expanse, and at least initially, without all, any certainty that there's land on the other end, right? So, that's kind of how I view our space program is like-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... big, you know, very conscious deliberate ef- efforts to get from point A to point B. If you look at how Pacific Islanders, um, transmitted, you know, their descendants and their culture and so forth, throughout Polynesia and beyond, it was much more, you know, inhabiting a place, getting to the point where there were people who were ambitious or unwelcome enough to decide it's time to go off-island and find the next one, and pray to find the next one.
- LFLex Fridman
Mm-hmm.
- RRRob Reid
That method of transmission didn't happen in a single con- swift year, but it happened over many, many centuries, and it was, like, going from this island to that island, and probably for every expedition that went out to seek another island and actually lucked out and found one, God knows how many were lost at sea. But that form of transmission took place over a very long period of time. And I could see us, you know, perhaps, you know, going from the inner solar system to the outer solar system, to the Kuiper Belt, to the Oort cloud. You know, there's, there's theories that there might be, you know, planets out there that are not anchored to stars, like-
- LFLex Fridman
Mm-hmm.
- RRRob Reid
... kind of hop, hop, slowly transmitting ourselves. So, at some point, we're actually in a- Alpha Centauri. But I think that kind of backup copy and transmission of our physical presence and our culture to a diversity of, you know, extraterrestrial, um, outposts, is a really exciting idea.
- LFLex Fridman
I really never thought about that because I'd, I have thought... My thinking about space exploration has been very Atlantic Ocean-centric in a sense that there would be one program with NASA and maybe private, uh, Elon Musk's SpaceX or Jeff Bezos and so on. But it's true that with the help of Elon Musk making it cheaper and cheaper and more effective to create these technologies where you could, uh, go into deep space, perhaps the way we actually colonize the solar system and, um, and expand out into the galaxy is basically just, like, these, like, renegade ships of, of, uh, weirdos (laughs) -
- RRRob Reid
(laughs)
- LFLex Fridman
... that just kinda, like, like, ho- like, most of them, like, quote/unquote "homemade," uh, but they just kind of venture out into space. And just like, like, uh, you know, the Android d- the initial Android model of, like, millions of, like, these little ships just flying out, most of them die off, uh, in horrible accidents, but some of them will s- will persist or there'll, there'll be stories of them persisting. And over a period of decades and centuries, there'll be o- other attempts almost always as a response to the main set of efforts. That's interesting.
- RRRob Reid
Yeah.
- LFLex Fridman
'Cause y- you kinda think of Mars colonization as the big NASA Elon Musk effort of a big colony, but maybe the successful one would be, you know, like a decade after that, there'll be, like, a ship from, like, s- some kid, some high school kid who gets together a large team and does something probably illegal and launches something where they end up actually persisting quite a bit. And from that, learning lessons that, uh, nobody ever gave permission for, but somehow actually flourish. And, and then take that into the scale of, um...... centuries forward-
- RRRob Reid
Mm-hmm.
- LFLex Fridman
... into the, into the rest of space. That's really interesting.
- RRRob Reid
Yeah, I think, I think the giant steps are likely to be NASA-like efforts.
- LFLex Fridman
Right.
- RRRob Reid
Like, there is no intermediate rock, well, I guess, there's the moon, but even getting to the moon ain't that easy between us and Mars, right? So, like, the giant sa- steps, the, the big hubs, like the O'Hare airports-
- LFLex Fridman
Yeah.
Episode duration: 2:59:12
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode cuD9uNFXnU8
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome