Making Sense - Sam Harris - #463 — Privatizing the Apocalypse

00:00:00.000 Welcome to the Making Sense Podcast. This is Sam Harris. Just a note to say that if you're hearing

00:00:11.980 this, you're not currently on our subscriber feed, and we'll only be hearing the first part

00:00:16.320 of this conversation. In order to access full episodes of the Making Sense Podcast,

00:00:20.860 you'll need to subscribe at samharris.org. We don't run ads on the podcast, and therefore it's

00:00:26.400 made possible entirely through the support of our subscribers. So if you enjoy what we're doing

00:00:30.400 here, please consider becoming one. Okay, Rob Reed, thanks for coming back on the podcast.

00:00:39.680 It's good to be back. So we've done, yeah, you probably have a better count of the number of

00:00:44.280 podcasts we've done on this topic than I do. I mean, we did one that was a very deep dive that

00:00:48.540 was, you know, more highly produced where it was almost like your audio book framed by our podcast

00:00:53.540 conversation but we share this concern around biosecurity and pandemic risk and bioterrorism

00:01:00.180 uh and you have an update for us on um the fate of the deep vision project yeah but before we jump

00:01:07.220 into that just remind people how you got to this topic what do you um how did you come to be focused

00:01:12.180 on this and and how much of your bandwidth has it taken yeah yeah well my my full-time job is in

00:01:18.180 venture capital i run a fund that invests in companies that we think will make the world

00:01:21.940 world more resilient in some important way. Fabulous job. I do it with a gentleman that

00:01:26.640 you know very well, Chris Anderson of TED fame. So that's my full-time job. In this case, I have

00:01:32.520 been, I guess my public service, you know, side of life, voluntary side of life has been focused

00:01:38.200 entirely on bio-risk for about a decade. It started when I was writing a sci-fi novel called

00:01:43.180 After On, and that had a subplot in it about a nihilistic kind of cult that thought it would

00:01:50.460 please god tremendously if they killed every person on earth it wasn't the center of the book

00:01:54.440 and so they use synthetic biology which i'll abbreviate to synbio just to make you save us

00:01:58.680 some syllables yeah to come up with a an omnicidal pathogen that could hopefully do that and that 1.00

00:02:04.540 started me worrying about this particular category of risk and you gave a ted talk yeah that came

00:02:10.580 that there was a couple dominoes later so i um i started worrying about this category of risk i

00:02:15.180 interviewed scientists in order to write this book accurately and then um i started a podcast called

00:02:19.820 after on same title. And I explored the topic there, including an interview with a brilliant

00:02:24.860 person that I'm sure a lot of your listeners know, Naval Ravikant. And we talked about that. And that

00:02:30.100 was about 10 days before the TED conference. So the TED folks called me up, said, would you like

00:02:34.380 to do a talk about this? Usually people have several months. More than 10 days. Yeah, more

00:02:38.680 than 10, but it went well. And you first entered the picture at that point because you were at TED,

00:02:44.880 you came up to me and said, hey, we should do something ambitious on this important topic.

00:02:49.300 as podcasters, you know, maybe team up. And then along comes COVID a few months later. And I really

00:02:54.540 rabbit holed into the topics of synbio risk and pandemic resilience. And you and I did end up

00:03:00.300 doing that magnificently sprawling almost four hour episode on the subject. That traveled far

00:03:06.440 and wide. And eventually somebody from the White House reached out to me, White House staff,

00:03:11.400 and said, would you like to come in and present to some, you know, pretty senior people in

00:03:16.240 biosecurity so i did that i went to washington and it was through that i that i learned about

00:03:20.980 this crazy program called deep vision which had just been authorized it was growing up inside of

00:03:26.260 usaid of all places it had 125 million dollar five-year budget and in the words of one very

00:03:32.460 wise person in the field of biosecurity it had in a worst case the potential to cancel civilization

00:03:39.680 is how it was put to me and with the best of intentions with the best we actually literally

00:03:44.780 with the best of intentions, but what a terrible thing to do, right? So I learned about it and

00:03:49.380 decided that that may not have been an exaggeration and decided to do my best to blow the

00:03:55.040 whistle on it. So I actually called you at that point and I told you what was going on.

00:04:00.140 And I told you that I thought the best way to blow the whistle on this would be to have a

00:04:03.760 really extensive interview with a professor at MIT named Kevin Esfeld, who I would characterize

00:04:09.920 I think is an evolutionary engineer, and he's very, very deep in this program. And you made

00:04:15.160 the suggestion, which was an excellent one, that I should interview Kevin because I was pretty deep

00:04:19.420 in the subject already, and that we could both, I'd create an episode of my podcast, which we

00:04:24.040 could both then broadcast to our audiences, with yours being much, much larger in hopes that

00:04:28.680 somebody would hear it and help to do something. So we did just that, you and I. And it was actually,

00:04:37.740 You might remember, I'm sure you remember this. You and I had an audience of one in mind, which gave us optimism that this might work, which was Samantha Power. She was running USAID at the time. And I think her husband had just been on your podcast and I had a couple people in common with her. But unhappy accident of history, I think just days before we posted this episode, the Ukraine invasion happened and USAID was very busy there. So it was actually a couple of months of crickets, but then things started to happen.

00:05:07.540 Yeah, yeah, yeah. So let's remind people, first of all, that deep dive is still in our podcast feeds to be listened to should anyone want to hear it. Because we go into just the larger set of concerns around, you know, SynBio and pandemic risk. But let's focus on deep vision. What was deep vision and what happened to it?

00:05:29.520 It was three really bad ideas, arguably each one worse than the one that came before. So Deep Vision was going to do three things. The first one is called virus hunting. And virus hunting basically, in this context, was going to involve going out to a dozen developing countries where they're going to be doing business. I think they wanted five in Africa, five in Asia, two in Latin America.

00:05:49.460 and going to very remote places like bushmeat markets, isolated bat caves was going to be a

00:05:54.920 very, very big one, and tried to discover roughly 10,000 undiscovered viruses of unknown deadliness

00:06:02.300 and extract them from these remote places and bring them into very leaky, imperfect vessels

00:06:09.180 in dense population centers called laboratories. And I categorize laboratories that way because

00:06:15.380 every category of laboratory all the way up to the highest biosecurity level demonstrably leaks.

00:06:21.740 There's plenty of history that shows that. And the alarming thing is we do not know the rate at

00:06:26.160 which they leak because there is no uniform reporting system, et cetera. We just know that

00:06:31.780 they do. And they, in some cases leak prodigiously, which means that an isolated bat cave that nobody 1.00

00:06:37.300 is otherwise ever going to enter is a much better, safer place for a pandemic grade pathogen

00:06:42.240 than a lab that's staffed by imperfect humans.

00:06:45.540 This has been a longstanding practice, though.

00:06:48.320 The virus hunting was a thing, I remember,

00:06:50.700 before I ever heard of Deep Vision or this specific project.

00:06:54.160 And it seemed a sensible practice on its face.

00:06:58.180 I mean, it wasn't obvious what was wrong with it.

00:07:00.720 And probably still isn't obvious to many virologists

00:07:04.000 who are incentivized to not recognize that it's a problem.

00:07:08.320 What is the stated motive for going into caves and sampling from the virome of bats and bringing

00:07:16.000 that out into the open? Well, I'll tie it to the next goal of deep vision, which was going to be

00:07:21.380 characterization, which is a series of four experiments that would determine which of these

00:07:26.000 viruses were most likely to be true weapons of mass destruction, most likely to be pandemic-grade

00:07:31.760 viruses. Why would you do those two things? In theory, it seems to make great sense

00:07:38.240 that you would want to find out what the pandemic-grade viruses are and where they're living

00:07:41.760 so you can start monitoring the interfaces between the human population and where those

00:07:46.600 viruses are living. The fact is you can do that kind of monitoring very, very robustly with

00:07:53.220 traditional public health methods. And the danger that happens is if you find these viruses,

00:07:58.900 you extract them, you bring them into places, into leak-prone laboratories, and then you do

00:08:05.040 this characterization work and you find out like, wow, these are profoundly deadly things,

00:08:11.880 there's not a lot you can actually do with that. You can't make a vaccine, for instance,

00:08:17.420 because the way that we know if vaccines work is we wait for there to be an outbreak where we start

00:08:22.480 inoculating people and discovering whether the inoculated people are healthier than the

00:08:28.140 un-inoculated people. If you did this hypothetical act and you found a deadly virus and you determined

00:08:35.520 that it was really, really dangerous, quite possibly a pandemic virus, you might come up

00:08:39.980 with a vaccine candidate, but you're not going to have any knowledge of safety or efficacy.

00:08:45.160 And because you're not going to infect a bunch of, you know, healthy volunteers with a potentially

00:08:50.880 deadly virus in hopes that the half of them who are not in the control arm and get the vaccine,

00:08:56.620 maybe the vaccine works, it doesn't work, you will have the vaccine candidate. And so that's

00:09:02.120 not useful knowledge. And it's actually very, very damaging knowledge because if it becomes

00:09:07.760 widely known that this pathogen might be a real doozy, it's going to become the most famous

00:09:12.900 pathogen on the planet or one of them. And the next thing you know, maybe dozens or even hundreds

00:09:18.340 of laboratories are studying it in BSL-2 or BSL-3 labs because it wouldn't be in a BSL-4. And these

00:09:24.580 the gradations of biosecurity because it's of unknown deadliness and that tends to push it to

00:09:30.340 BSL 2 or 3. And so now you potentially have this dangerous thing that's being studied throughout

00:09:36.540 the world. And, you know, for anybody who believes that there's a significant probability that the

00:09:42.160 Wuhan virus was a leak, it becomes self-evident that you don't want these things being studied

00:09:47.200 broadly. Now, the third thing that deep vision was going to do was to me the most objectively

00:09:52.480 crazy one, which was having found these 10,000-ish viruses and established which ones were the most

00:09:58.540 likely to be truly deadly, they were going to publish that list and also the genomes of these

00:10:03.760 viruses to the entire world. Isn't that helpful? Isn't that helpful? A world, which it's important

00:10:09.840 to point out, containing at the time roughly 30,000 people, according to Kevin's best estimate

00:10:15.980 at the time who had the tools and the know-how and the wherewithal to then conjure those

00:10:22.520 viruses, basically make them from scratch using techniques that are called reverse genetics

00:10:27.160 and sometimes it's called viral rescue, but about 30,000 people.

00:10:31.400 And so what this meant is you were potentially giving the killing power of a nuclear arsenal

00:10:37.160 to 30,000 completely unvetted strangers throughout the world, some of them almost inevitably

00:10:42.500 located in islands of stability like pakistan north korea iran etc yeah yeah and uh i mean you

00:10:49.360 just look at the the mental health uh probabilities over any population i mean just if only one percent

00:10:55.500 of them uh had brains ready to go haywire uh with those skills it's an abiding problem and uh that

00:11:01.940 was pre-ai we're going to talk about the contributions that ai is making to this issue

00:11:07.700 Okay, so we made a bunch of noise about Deep Vision.

00:11:10.660 Yeah.

00:11:10.920 So then what happened?

00:11:12.380 What happened was the following.

00:11:13.800 So you and I were hoping to influence Samantha.

00:11:17.180 We didn't succeed in that.

00:11:18.800 Ukraine invasion, great deal of distraction.

00:11:21.600 Couple months of crickets.

00:11:23.200 And then a friend of mine reached out to me, Tristan Harris, who's also been a guest on your show.

00:11:29.040 Yeah.

00:11:29.280 Along with...

00:11:30.280 Also, if memory serves, forced to give a TED Talk on like a 72 hours notice or something.

00:11:36.620 Yeah, he's really good at that. I think he had no time at all. Yeah. No, he beat my record and he seemed so smooth in practice. He called me along with a person named Daniel Schmachtenberger. Do you know Daniel? Yeah. Very, very interesting thinker. Thinks a lot about existential risk. And so the two of them called me and we talked, you know, over Zoom. This is still kind of mid COVID about this risk.

00:12:00.840 And then Daniel curated a group of, I want to say, seven or eight people, really brilliant folks. And then he and Tristan hosted it quite close to my home in Northern California. And we had what was, you know, kind of like a 12 or 13 hour brainstorm, a really electrifying conversation.

00:12:19.980 And people from, you know, who are experts in bio, people who are experts in existential risk.

00:12:26.740 Our friend Lavori was there, a few other people.

00:12:30.120 And as a result of that, Daniel decided that he was going to really run with the ball.

00:12:34.800 So he got out and he runs an organization that thinks very deeply about existential risk.

00:12:40.640 He also knows far more people than I'll ever know in Washington.

00:12:44.180 And so he started reaching out to folks and he soon reached an organization called Helena.

00:12:49.980 Yeah, I know those guys.

00:12:50.700 Yeah, brilliant, brilliant, interesting group of people.

00:12:53.940 They're basically a problem-solving organization.

00:12:56.900 It wouldn't be accurate to call them a think tank.

00:12:59.060 And what they'll do is they'll identify global problems.

00:13:02.720 They'll spin up groups of kind of cross-disciplinary experts to try to solve them.

00:13:07.640 Sometimes they'll start, you know, a not-for-profit to tackle the problem.

00:13:11.080 But they also have an investment arm that invests in world-positive companies, which is my day job.

00:13:16.940 So we have a lot in common.

00:13:17.760 There was somebody there at Helena named Prodick Basu, who in particular spent a lot of time in Washington and knew a lot of folks. And so Prodick started sniffing around and he very quickly found out that a couple people were already on the case, specifically Lindsey Graham and Senator James Risch of Idaho.

00:13:35.180 They and their staffs had become aware of Deep Vision, and I think they'd sent two letters

00:13:39.440 already that were mainly about other topics, but expressed concern about Deep Vision to USAID.

00:13:44.460 So there's already this tremendous pressure coming from the Hill.

00:13:48.180 And then Prodick started looking out for other people he could bring into sort of a loose

00:13:52.660 alliance.

00:13:53.620 He knows people in both parties, people inside the administration, outside of the administration.

00:13:59.020 Inside of the administration, naturally, you're going to tend to be Democrats.

00:14:02.060 He reached out to people inside of USAID, people in security.

00:14:05.980 One of the people that he reached out to who was very helpful was Chelsea Clinton, who

00:14:10.500 was, you know, really, really helpful in reaching out to a lot of folks.

00:14:13.100 She has a master's in public health, a big network there.

00:14:15.680 But it became a very, I'd say, extremely unpartisan group.

00:14:21.040 Also, Rand Paul, I think, was significantly important because he had a hearing the summer

00:14:27.220 of 2022 that Kevin Esfeld testified at talked about deep vision. And so this is a pretty,

00:14:33.300 you know, unpartisan group. You look at the spectrum of folks. And what happened was just

00:14:39.380 a lot of people started working very quietly, very quietly for the most part, to try to put

00:14:44.260 pressure on this thing. And we learned probably a couple months after, you know, Daniel and Proto

00:14:51.360 got heavily involved, that there had effectively, the program had effectively been defanged. That

00:14:57.220 Due to the pressure coming in from all these different points, there was not, no work was going to be done.

00:15:03.760 It would never be done.

00:15:05.200 And the problem was more or less passed.

00:15:07.640 And that was great news and took it at face value.

00:15:10.640 But then intriguingly and unexpectedly, about a year later, the program was formally killed.

00:15:16.440 And we didn't expect that to happen because that would be egg on various faces and so forth.

00:15:21.320 But this was also a really, really positive thing because the public demise of this program, you know, sent a pretty strong signal that we don't do this sort of thing anymore, hopefully. And, you know, I think, you know, the pressure continued from Capitol Hill.

00:15:36.180 I know that James Risch sent a letter as late as May, very anti-deep vision letter, probably the

00:15:42.380 third to USAID. And it was September of 23 that we found that this thing was ultimately and

00:15:50.200 completely killed. And with that, I would say an enormous source of plausible risk exited the

00:15:58.340 equation. What about other countries doing that same work? Were we the only ones playing this

00:16:03.280 game or, or do other people go into bat caves and other, uh, you know, go hunting for vectors of

00:16:09.740 awfulness? Yeah. I think that the Wuhan Institute of Virology has been doing this for a very long

00:16:14.620 time. So WIV very heavily into, you know, collecting and understanding coronaviruses,

00:16:20.980 USAID had previously funded a program called predict, which did this at a pretty big scale.

00:16:26.700 I think they've, they discovered something like 1200 novel mammalian viruses, but I don't believe

00:16:32.580 that there was ever a virus hunting program anywhere near the scale of Deep Visions.

00:16:39.120 And the interesting thing when you think about the risk landscape is to contemplate how it's

00:16:44.740 changed from 2021 when Deep Vision was authorized to today. So you go back to that period of time

00:16:51.740 and it's remarkable how few entities were in a position to have an idea, frankly, this bad,

00:16:59.860 right? In its worst case, and we can talk about why, and it's probably valuable to you in a

00:17:04.220 moment, but it is, you know, deep vision had a clear potential to cause death at the scale of

00:17:09.420 COVID. Not definitely, but it certainly had that potential and possibly far, far worse.

00:17:15.000 Probably worse. Yeah. Probably far, far worse.

00:17:17.080 In the scheme of things, COVID was remarkably benign as an infectious agent, right? I mean,

00:17:23.340 it was super infectious, but it was not super lethal, right?

00:17:26.020 Very far from super lethal, yeah.

00:17:27.680 So it could have, I mean, again, I really do think of it as a dress rehearsal for something awful, and we appear to have failed this dress rehearsal in a variety of ways.

00:17:37.500 Oh, we botched it spectacularly.

00:17:39.260 There's no question about it.

00:17:40.280 But think about COVID, like Deep Vision, and we may or may not get into the numbers, but a conservative estimate is that they may easily have found, you know, six, seven, eight pandemic-grade viruses.

00:17:51.420 Now imagine a really malevolent actor like an Aum Shinri-kyu deciding that we're going to really delight the heavens if we take down civilization. 0.99

00:18:02.540 COVID itself emitted from one single point, and it approached our shores at a speed of four and a half miles per hour.

00:18:10.720 That's the back of the envelope.

00:18:12.220 It took two months to get here.

00:18:13.720 Two months to brace ourselves for that, and obviously it knocked us to our knees and the rest of the world.

00:18:18.820 Imagine seven pathogens emerging all at once from 20 different airports, a complete worst-case scenario. I don't know how we survive that. The combined fatality rate could certainly be way beyond COVID. Doctors would have no idea which of these pandemics they're diagnosing. People could be afflicted with more than one at the same time.

00:18:43.380 that's the situation where you don't worry about civilization toppling necessarily because

00:18:48.860 everybody gets infected but you do if you get to a point where no thinking frontline worker

00:18:54.720 is going to go out the door and risk killing themselves and their whole family

00:18:58.520 for gig worker wages and and when that happens the supply of food law enforcement eventually 0.96

00:19:04.640 electricity and everything else shuts down and so that is a profoundly profoundly risky scenario so

00:19:11.420 anyway, back to Deep Vision 2021. It's amazing how few people could have thought of an idea

00:19:16.920 with this level of potential destruction. Definitely not terrorists. I mean, Osama

00:19:21.940 bin Laden himself never had a wisp of that potential destruction. Not the world's worst 0.81

00:19:27.260 criminal gangs or cartels. They only have conventional weapons. I mean, not even a rogue

00:19:31.580 state as gigantic and chaotic as Iran could have dreamt a killing at the scale of COVID. And so 0.98

00:19:38.080 you're basically left with nuclear weapons, nine people in that category, I guess, and biology.

00:19:44.380 And in the world of biology, it's amazing to think of how few entities had the capability of

00:19:51.280 marshalling budgets as large as Deep Visions, $125 million, and on top of that, access to

00:19:57.560 scientists, expensive labs, and to forge partnerships in a dozen developing countries

00:20:02.820 in which they were going to recruit scientists to, you know, find, you know, lots of viruses

00:20:08.360 and poke at them. I doubt if even 10 entities in the world could have come up with an idea,

00:20:14.740 let alone implemented that in 2021. And the remarkable thing and the very important thing is

00:20:20.820 somehow one of them did. And then think about the people who...

00:20:26.200 But again, with the best of intentions, I mean, somehow they're missing the fact that this is

00:20:30.040 raising risk of accidents or, you know, malicious use that's not intended by the people framing the

00:20:37.840 project. It's just, again, you don't know how many other ideas are this bad and not acknowledged to

00:20:43.240 be this bad, but it's quite amazing to be blind to the downside of this effort.

00:20:49.220 And the best of intentions is the other side of this. Like, I have no idea who was on the USAID

00:20:54.340 committee that came up with this idea, but I am quite confident it included no mass murderers,

00:20:59.060 terrorists or dictators, without any questions. So somehow a very, very, very tiny population

00:21:04.980 of well-placed, highly-placed do-gooders came up with this idea. So that is a very powerful

00:21:11.660 and grounding lens through which to look at a coming era, a near-term era, in which untold

00:21:18.240 thousands of people will be empowered to have ideas as bad as deep vision and worse.

00:21:23.740 And some of them will be terrorist mass workers.

00:21:25.520 if you'd like to continue listening to this conversation you'll need to subscribe at

00:21:31.080 samharris.org once you do you'll get access to all full-length episodes of the making sense

00:21:36.380 podcast the making sense podcast is ad-free and relies entirely on listener support and

00:21:42.520 you can subscribe now at samharris.org

00:21:55.520 Thank you.

Making Sense - Sam Harris - March 11, 2026

#463 — Privatizing the Apocalypse

Episode Stats

Summary

Transcript