"DEEP VZN" Revisited | Kevin Esvelt

Rob Reid (0:04)

Welcome to The After On Podcast. I'm your host, Rob Reid, bringing you conversations with thinkers, founders and scientists. They're in-depth and unhurried and are structured to bring you up to a top percentile understanding of something important. So whether you're into startups or ideas, a techie or a lit major, take your time, engage your mind, and you'll be glad you did.

This is a reposting of an episode I first ran way back in 2022 It's an interview with MIT's Kevin Esvelt about an insanely misguided virology research program called Deep Vzn. USAID spun up Deep Vzn with $125 million budget, and it had insane destructive potential. Had it delivered on its objectives, I find it impossible to imagine a scenario in which it wouldn't have triggered untold millions of deaths. These awful consequences may or may not have come quickly, but they definitely would have come over time. This interview was my attempt to blow the whistle on Deep Vzn. I was greatly assisted in this by Sam Harris, who also posted my episode to his much larger audience. I'm about to appear on Sam's podcast again to discuss Deep Vzn and how it was shut down. And above all, the lessons we urgently need to learn from this outlandish near-miss as we approach a much more fractured and alarming risk landscape than we faced back then, driven by the collision of AI and synthetic biology. So I thought it would be useful to repost this interview that touched everything off. As a context for my own audience, who may want to check out my appearance on Sam's podcast, and maybe for some of Sam's listeners who might find their way here to get the full Deep Vzn story. Back to the history. After Sam and I first posted this interview with Kevin, some people reached out to me who, unlike Kevin, me or Sam, knew how to navigate Washington. They discovered that some influential people were already sniffing around Deep Vzn. Then they quickly assembled a remarkably diverse alliance, both politically and in terms of professional backgrounds. Eventually, Deep Vzn was de-fanged before it could do any damage. And quite some time after that, it was fully and formally terminated. No one person took the kill shot. Instead, lots of pressure emitting from lots of different angles ended the program in those two distinct phases. I didn't know a lot of this story myself until I discussed it with some people to prepare for my appearance on Sam's show. And I still don't know all of it. But I can say it's an inspiringly unpartisan story of people from across the political spectrum working quietly and without seeking credit to achieve something simply because it was urgent. And again, the most important aspect of the story is what it teaches us about the risk landscape we're now entering, which is very different from the one we faced when Deep Vzn was authorized in 2021, after which it could have easily been as deadly as COVID and possibly much worse. What's truly remarkable, though, is how few people were in a position to have an idea this bad back in 2021 Not terrorists. Osama bin Laden himself never had a wisp of this destructive potential. Not the world's biggest criminal gangs or cartels. They only have conventional weapons. Not even a rogue state as giant and evil as Iran could have dreamt of killing at the scale of COVID. So if we exempt nuclear weapons, which are not a new feature on the world stage, this pretty much leaves biology. And in that world, very few bodies had USAID's ability to marshal Deep Vzn's budget, its access to scientists, labs and expensive equipment, or to create partnerships in the dozen developing countries where Deep Vzn planned to unearth and poke at deadly undiscovered viruses and to train local scientists so they could keep doing this on their own indefinitely after the program ended. Yeah, really. I doubt if even a dozen entities worldwide could have authorized a program of Deep Vzn's ambition, scale and flat out lunacy in 2021 And somehow, one of them did. Now, I have no idea who was on the USAID committee that dreamt up Deep Vzn, but I'm confident that it included no mass murderers, terrorists or dictators. In fact, I'm sure they were all perfectly good people who didn't intend to hurt anyone. Fast forward to sometime later this decade. We're about to enter a world in which many thousands of people will be empowered to act on ideas as bad as Deep Vzn or worse. And some of them will be mass murderers, terrorists and dictators. This is due to accelerating developments in synthetic biology which had been racing ahead on their own for years and are now intersecting with the awesome capabilities of AI. As this unfolds, the barriers of expertise, budget and permission needed to do things that only the world's top biologists can pull off today will plummet. And plummeting barriers mean proliferation. I'm sure we'll be discussing this on Sam's show, which I expect to appear as soon as March 11th and definitely before the end of the month. But for now, you can hear the full range of dangers that Deep Vzn almost injected into the world on behalf of all 8 billion of us. And the key takeaway is not so much the details of all that, but the fact that a minuscule population of highly placed do-gooders improbably spawned this awful idea and came within a whisker of implementing it. This is the most powerful and grounded and realistic lens I can think of to look through at a near future in which this ability proliferates. And this is not a message of despair because there is a lot we can do to protect ourselves from the future I just described, which I'm sure we'll discuss on Sam's show and at some point here as well. I also plan to write about it on my Substack. Speaking of which, I am working on a piece about Deep Vzn to post at robreed.substack.com. It will hopefully be there by the time you hear this and if not, within a couple of days. And with that, I bring you the Deep Vzn story. Don't let it give you nightmares because we killed this sucker. And while the road ahead is sketchy, we still have time to take evasive action.