d/acc closing panel | Devcon SEA

Vitalik, Juan, Eli, and Robin Hanson. Thank you. And while they get seated, we've got two awards to give out for those who were very attentive. Firstly, whose brain was scanned and was on stage earlier today? Okay, I've got a hand over there.

Yes, coming up to the stage and get your gift. All right, the second one I have is whose talk actually spoke about these awesome air quality devices? What was the name of the person whose talk it was? Unbelievable. No.

No. Yes. Yes, come up to the stage and grab your device. And on that, we'll get started with the our panelists. Thank you.

Awesome. Yeah, I'm so excited to kick off this panel after such an amazing day, and looking forward to have a fun chat. So, I want to kick it off with a high-level question also for the people maybe that didn't see the whole day about kind of like what DEK is in in almost like differentiation to some of the other movements that came before it from effective altruism, accelerationism, like transhumanism, all of those different areas. Maybe Vitalik, you want to kick it off giving kind of like your definition of like almost like which movements also inspired you potentially and like how you yeah, came to some of those ideas with DEK. Yeah, so when I was started writing my post on techno-optimism, I think there were a few different things going through my head at the same time, right?

One of them was seeing the discourse on techno-optimism, and of course Marc Andreessen had published his famous EAK manifesto about a few months uh before that, and the debate between that and AI safety people that was like very very strong at that time. And um also I think just uh an effort to like even make clear to myself as much as I make clear to other people and of what even like are the principles that I'm that I'm like seeking to try to achieve and uh move toward and like what is the thing that's like updated to the to the 21st century that uh Um yeah, because they they at the time you know crypto I think has been for a long time following these ideas that originally came around in like 2008 and 2013. I think a lot of the core thoughts were based on you know January uh you know 2009 uh Satoshi or uh central bank on on brink of second bailout for banks and like all of that stuff, right? And like those ideas were at the time already 15 years old and they were very different from things that at any like other points in our our lives over the last few years like we've actually been concerned about. And so the goal was to try to like actually create an updated philosophy that would bring a lot of those thoughts together to as well as try to make a more concrete picture of uh like what kind of productive role crypto could play in this kind of future.

And so that's where the post was born and the Diac itself is basically this uh attempt to find I mean I don't want to just say find the middle ground because like finding the middle ground is like such a cliche, right? And like the whole concept of like you know like the egg the the third thing is like so overused that there's like literally it's like there's this concept of the third way and there's this concept of the third position and if you look at what they actually mean they're like almost opposites of each other. I think it's uh in this case an attempt to like integrate the parts of both that make a lot that that I think make the most sense and like try to create something that is like very philosophically coherent in its own way but that's still like actually brings the kind of future that we want. If any one of you want to add some perspective on your thoughts? Maybe I'll add that I think in particular the EA EA debate created a whole bunch of confusion in the world that I think EA is solving.

A lot of the EA philosophy initially was extremely accelerationist was very oriented towards progress and accelerating good positive change and crucially EA took on a defensive a selective acceleration path of their specific technologies that you want to accelerate first and this is an idea that a lot of people pioneered before EA as well so these these ideas are older but then EA specifically took a very strong position on just going all the way accelerate everything you know minimize um just just just accelerate the the path of minimize sort What what is the right phrase like minimizing pre-entropy? Uh some something like this where it's just like trying to organize try to let matter and computation go as fast as possible is totally fine if AI you know grows and obliterates humanity and to minimize uh reorganize the world into paper clips. And this was a core of the This is not like a fringe belief in E Act that was kind of a mistake. It was just a core component. Um and so a lot of the people that think about E Act don't know that.

Um and so the synthesis between EA and E and E Act is actually quite large. Like the two communities agree on way more than those communities actually think. Um and they I think of the act as a way to um leverage a lot of the good parts of those communities, leave that set of mis- of disagreements um kind of behind, and say there's a tremendous amount of positive outcomes that are possible, but it really does matter which things we get first. And it really does matter how you orient the outcomes. You could orient them towards um bad things or good things, and we have agency in that choice.

And so to the extent possible, let's try and and uh minimize the bad outcomes. So it's not a a try to just accelerate for acceleration's sake. Um and it's not a hey, halt all progress uh sort of perspective, which both of those things are now caricatures of those movements. I'm not trying to say that those that's what they mean. Um and just instead try to find a much more nuanced perspective of finding the right path, uh accelerate progress, but don't wipe yourself out and and kill the species while doing so.

Yeah, I mean one brief thing I would also add is that I think the EA versus E Act frame, at least as it existed a year ago, also created the impression that like whether it's a hero or a villain, AI and AI developers are like the only main character that matters, and uh San Francisco is the only city that matters, and uh like focusing on a really small number of narrow things. And uh to me a big part of uh DE Act is also creating a much bigger tent of people working on all kinds of technologies and all kinds of environments that we've seen presented today. So I think like kudos to the EA people because like this is all a reaction to sort of to something that they kind of kicked off, right? So, I think of EA as like very like seriously applied like Oxford utilitarianism, right? Like the the work coming out of like Peter Singer and and Derek Parfit.

And and I do think that it does have a blind spot, which is it's like very top-down and it's very um strongly convinced of the truth of utilitarianism as like a moral perspective. And um and I I I think like the EAC reaction is like mostly vibes uh based. It's like it's not uh it's not similarly rigorous. Um and I think that uh DEAC has is a much more reasoned reaction to EA based on pluralism and based on the idea of like, okay, maybe like um you know, top-down utilitarianism, you use to like decide where to put a bridge or something like that. But in in ordinary human life, there's many other values that we need to consider including the values of of of freedom and human rights and and other things that we wouldn't um we wouldn't like broker a a uh like a global totalitarian state to like stop uh you know, a a a hypothetical risk or something like that.

So, um so, I I see it as I see the one of the the values of the DEAC response to EA is like, no, let's let's have a more pluralistic uh response. And what I wonder is almost like as an observation for me is like talking into a lot of people like that that consider themselves EA or like EAC. I think funnily enough like a lot of them have more actually like a DEAC position, but it's it's just like the the meme uh is very polarizing, right? So, they're like get stamped as like this extreme safetyist or these sales or this extreme accelerationist. But then if you talk to them like to a lot of the scientists or researchers or startup people, like they actually have uh this more like reasonable middle ground position.

So, I wonder like and I think that there's like a strength to the meme and like to the the polarization of like EA or EAC that say progress studies, DEAC, and like other such movements are missing. Like they're not as like edgy, right? Like they're more reasonable. Like almost maybe too reasonable. So, I wonder like what your perspective is on this.

I was like, is that also like a risk to take like the synthesis of like two polarizing views um or like basically it's like how to what's for DEAC and those movements to succeed in that like memetic warfare almost. So, I got my PhD in formal political theory and that contains this essential contradiction that sort of as a mathematical game theory modelist from STEM, we're all about precision in language and being really clear about what we say, and then this whole world of slogans and you know, teams and all this sort of vague flags is just really hard to navigate. There's a basic fact that every one of these labels is vague. And then the it's each sort of pulled and pushed by its members and its opponents to an extreme to be a stark thing that everybody else can criticize easily or understand easily. And then as Vitalik said, there's always this temptation of saying, "Well, I'm the reasonable middle.

I'm not those crazy extremists." But then the middle's kind of hard to say. So, I struggle here to like say what's the precise thing I could endorse with with a respectable STEM background and being trying to be precise. And I guess the thing I would most come to is the idea that of the D in DEACcelerate that uh you know, the EA people are often quite willing to be centralized and authoritarian in their recommendations. And the wise recommendation is say, "Fine, we share your goals.

We're going to be careful, but you're dealing you're you're invoking some pretty dangerous stuff there. And we think maybe we could do similar things in safer ways. Uh like I I more decentralized approaches a bit safer cuz there's the danger of invoking these centers and these authorities, and that's really the thing that I mean, that's kind of what puts off Anders really. He sees them invoking these centers, and he goes, "No way, no way. I I do not want that at all."

And he's trying to take the most extreme position he can to reject that, which then looks kind of silly on the other extreme. I'm curious like um we touched obviously on a lot of them throughout the day, but if there's specific areas that you think are most crucial almost like for differential acceleration and like for DAC to make progress on, if any one of you want to start on it. The big one that I think is perhaps most on my mind is neurotechnology. Uh and I think to me and I think there's a huge difference in opinion here on how far you think AI AGI might be or ASI might be. Um if you have different opinions on the time frame there, then you you probably think differently about other technologies.

Um I care about a lot of them uh a lot of the ones that discussed today. Um the one that I maybe represent more these days is neurotechnology because that's the field that is both um in its infancy to some degree, uh has is has way less energy, way less but way less um people and capital, and it's the one that I think needs to be accelerated the most from where we currently are in a defensive orientation. Um so that would be that would be my plug. It's a BCI and eventually whole brain emulation as a very large uh hedge to make sure hedge is the wrong word, but as a as a defensive push uh to make sure that when we do make AGI or ASI type um beings, we don't accidentally create a competitive species that wipes humanity out, um and instead we have a some kind of smooth upgrade or smooth synthesis um along the way. Yeah.

I mean, I think well, you know, I'm a a big fan of uh neurotech as well, but one other thing I would highlight is uh the uh cyber defense and uh info defense side of things. And I think there are the challenges basically that like threats that we uh get on the internet are rapidly becoming much more sophisticated, right? So, uh the as basically the bit one big switch that we've seen between like literally 2 years ago and today is that there's basically no longer any uh reliable way to tell that someone is human just by looking at their pattern of behavior on the internet. And that broke a lot of things, right? And uh you know, we're seeing uh all kinds of um new types of scams.

We're seeing all kinds of uh you know, like new uh type uh types of hacks. We're seeing um like all kinds of uh you know, like growing cyber attacks. Um like a growing ways to use these technologies politically. And uh like this the there is a risk that this basically becomes this frontier of like low-grade war um war of all against all that had that constantly risks escalating into a worse kind of war. And to me, the only uh way of uh resolving things like this without um you know, basically taking the Hobbesian solution, which basically says that there is uh one small group of experts that all trust each other that decide everything for everyone is to actually uh like figure out what is the uh defense um on that side.

And there are a lot of like really uh interesting solutions, right? So, like on the uh on the cyber side, like there is uh actually an opportunity to make formally verify a huge number of things that were not formally verified before. There's massive gains in specific types of computer security that have been made. There have been massive gains made on hardware, like more things happen on trusted chips. There's more work on making making hardware more secure.

There's uh all of these like new epistemic technologies that are being developed. Like how do we get to a world with like five more things that look like community notes? And like just seeing more experimentation in that space happen at a rate that is fast enough to keep up with the pace of change on the internet feels like it could be extremely good for us. And and for like our ability as a global civilization to just maintain cohesion as we enter this age of some form of AGI and then super intelligence coming to us. I agree with all of that and and relatedly I would say authenticated comms.

So like so we you know I think one thing that AI is going to be very good at is phishing and you know voice cloning and and doing doing you know calls on the phone and like our our communication systems are not ready for this, right? Like like our phone calls are completely like unauthenticated and our emails, you know, are easily spoofed and you know the the traditional problem that has plagued this is public key distribution and blockchains are actually maybe a great mechanism for distributing public keys and and you know we could we could have you know a system where you can be reasonably certain that the you know that you're talking to the right person even in a world where there's voice cloning and so I think that would be like something that is a near-term like very practical step but then also like in a world where you know, we do have maybe a malicious super intelligence like it wouldn't be able to exploit that that vector, right? And so it it is both a near-term benefit to to build something like that and and also a long-term you know, safety-enhancing. Like like I wonder almost like cutting to the the crux I think almost like between the egg and like some of the other movements. Like I would argue that a lot of them like actually share the majority in common like they they do all love rockets and like bio progress and progress across every vector.

I think it really comes down to turn extent AI and I'm I'm curious if like on your all's perspective, but like I really wonder if like what the position of the egg should be or like is on kind of like on AI development specifically because I think it's it's unclear almost like how to navigate it I think for a lot of camps and I wonder if there's because it obviously promises some of like like great like positives and upsides but also some potential risks and obviously both the closed source and the open source paradigm has risks and the centralized or the decentralized approach. So I wonder kind of like if you agree that this is like one of the core differences maybe also between some of those technological movements or if there's some others where the egg takes like a different position. A generation ago, I get to play the old man position here. Uh we had the great nanotech debate and I think we learned a lesson from that which was Eric Drexler tried very nicely to call attention of the world to an exciting technology and honestly paint both the upside and the downside but the net effect was people overwhelmingly looked at the downside and wanted to distance them themselves from it and assure them everybody that like that couldn't happen and they were just going to lock it down. And there's a plausible case for saying that the world kind of goes okay so far when the world doesn't really know what's coming and they bubble into it and they slowly get used to things, but when they see something a distance away and can describe it abstractly in ways that could, you know, terrify them, uh they don't react very well.

So, there's a basic problem here uh that maybe our world isn't very well set up for hearing ahead of time big things are going to happen. I I I've I feel there's the problem that a lot of us who know more about technology in the future, we want to identify with that knowledge and that difference and we want to really paint the fact that big things are coming soon and we know about and other people don't and hey you guys, you know, you need to wake up and hear, but the way most people react to that is oh, there's if, you know, if they actually believe you, cuz mostly they don't, if they actually believe you, they're going to react and say, "Oh my god, something big's happening. Something should be done. Someone should be in charge." And that's exactly the problem.

I kind of feel like it's different uh in maybe the last 10 years. I I I I think I I agree with a lot of the prior decades definitely operating that way and to a big degree some of the AI risk is being taken taken that way, but I do think in the last 10 years we've seen the effect of being able to talk about the future actually recruiting a dramatically large contingent of people to help build it and help accelerate the build out of that technology. Um and so I think these days, you know, I think it took a while for the internet to really percolate through a lot of the world, um but nowadays we are having gatherings like this in Thailand. We are not presently in a room in Silicon Valley talking about the future. We are now in um being able to gather people around the world on on a range of um a range of components.

And so, um I think we uh we have gotten the ability to marshal resources around the world. Well, there is a risk of scaring people and causing unnecessary or bad slowdowns. I guess uh, at the same time there's also risks to take into account. So, I think to like Vincent's um, prompt the I I I think it does I think it a lot of people's positions will really center on what they think will be the timelines on AI development and how quickly we'll achieve some things because some of those some of the things that you might try to achieve will feel not relevant to you uh, or in in a in a short enough time frame to have enough of an impact. Um, now I tend to take a very conservative position here in terms of um, broadly you should let as many parties as possible accelerate the range of things that they think because um, there's knowledge is not evenly distributed um, and is not equally distributed to everybody.

So, different people will be able to spot really valuable components and you should have a broad rights protecting system that lets people accelerate the things that they think are are valuable, but I do think at least for a set of us um, the the to to your AI specific question, um, I do think make doing something about that differentially would be valuable. Though, I do very much agree that you cannot quite slow it down and it's actually potentially even more dangerous to slow it down. Yeah, and I think I mean I'll like step out of the Kumbaya a little bit and like answer the original question which is what is a crux between uh, EAC and DAC. I would say uh, DAC is much more explicitly uncomfortable with acceleration of a type where the consequence of the acceleration is that one thing gets 51% of all of the EAC power of human civilization and regardless of whether that thing is a nation state or a AI philosopher god trying to implement a pivotal act, which was like one of the EA theories, or the first corporation that happens to figure out how to unlock the next scaling law. Um, and uh, like, I mean, to get into like a more concrete, like, I think, you know, the way that OpenAI originally started as this effort that basically said that, you know, for we uh, there are all these big corporations that are doing closed AI, and so we need something that is open as an alternative to make sure that these benefits widely or reach all of humanity.

And then, of course, as we know, 4 years later, they sacrificed openness for the sake of safety. And then, 4 years after that, they sacrificed safety for the sake of growing and getting even bigger. And like, honestly, that scares me. Uh, so, like, that that's like one explicit difference for me. I'll take like maybe just like a like a really agnostic approach.

I think like people are going to be People are I don't put a lot of credibility in like any of these forecasts, right? Like, I think that um, uh, even the experts, right? You said like like people, you know, like like lay people are not don't know what's coming. But like like even the experts are bad at predicting the effects of technology. And if you think about like I think about very often, um, uh, the Declaration of the Independence of Cyberspace, this document from 1996, John Perry Barlow, founder of EFF, also lyricist for the Grateful Dead.

Um, and and he thought he knew what was coming, which is that governments of flesh and steel were going to be obsolete, and we were going to create a society of mind in in cyberspace. And if I look at the internet today, I don't see a society of mind. And, you know, tell the people of Ukraine that governments of flesh and steel are irrelevant. It's not, right? And um, and so, even the experts don't know.

Um I you know, my my bias is the physical world still matters, not just the digital world. And um I do think you know, like if if if um if AI follows the pattern of the internet, it's going to be incredibly significant, maybe more significant, but a lot of that significance might be cultural or um you know, not in the not in the in sort of the physical like the physical world might be more complex than than a lot of people predict, and and it may be hard for AI to like uh affect it as much as we think it's going to. Let me agree with Eli so much he might disagree with me. If you look at uh say billion-dollar projects or $10 billion projects, they just consistently seem to go badly. And if you look at the history of regulation, I'd say the best regulations uh and regimes are ones where they you're responding to pretty specific problems with a pretty known context, and you have something that responds to that.

Uh and then, you know, if things change, you change that. And the worst cases in history, like things like nuclear power or even genetic engineering, are where you're responding to a pretty abstract concern that hasn't been very much realized, and people sort of make an abstract response to it that isn't you know, adapted to a particular context or how the problems that are realized. And that I think the lesson there is that we overestimate our ability to see and plan these things. And yes, you might envision various AI terrors, but honestly, so far hardly anything has happened. Let's emphasize this.

Hardly anybody has been hurt at all by AI, and we're not even close to anybody being hurt anytime soon by AI. So, I would say hold your concerns, watch for things, but wait till you see some actual harm in a concrete context and then we'll listen to your concerns about what we should do about that. But that's not a good way to avoid falling off cliffs. Well, if you have some light in front of you and you are if you get a few feet to the cliff you could stop maybe. I think I think it's precisely what we're talking about.

I think a few of us with think that we see the edge of the cliff and we're trying to avoid the edge of the cliff. Um and this is not going to smoothly fall on some slope. Um there there can be a very uh dramatic drop and I think that's what's fueling a lot of the the concern. So if we if we take your analogy of just wait until there's a bunch of damage, um I think that's up that's true up to a point um but but I think a lot of people argue um that may may very well be way too late. By the time you see substantial damage, um it might be uh it might be way too hard to recover.

I'm not necessarily you know in full agreement with that point, but I think a lot of a lot of people trying to selectively accelerate on things or decelerate AI would like take that position. In the history of technology, we just haven't seen cliffs of that sort. That's the maybe on territory like you know driving across a terrain you might come across a cliff but that's even really rare. Nuclear weapons we we blew up two and we said and then we tested a bunch and we said no more and they should not be used in war and I think today we are in a much better position because of that. But and we said a whole bunch of things go wrong if we made nuclear power and we made sure to stop those and we sort of killed the whole damn industry.

Yeah, yeah, I I very much agree with you on that part. Yeah, for sure. Right. So I guess the question is would the world be a better or worse if anyone could buy both a nuclear power plant and a nuclear Walmart? But plant and the uh But even the nuclear weapons was a response to an actually observed thing.

Nuclear bombs actually went off and we had a sense of how big they were and how expensive and how heavy and a whole bunch of details about nuclear weapons that we could respond to in regulating nuclear weapons. They weren't an abstract hypothetical thing that might happen in 20 years. To wrap things up, I would love to title back to Def Con 2 Ethereum and crypto. And I'm very curious like I think the audience is like largely builders in crypto. Like if I'm not mistaken, like a big chunk or people in the space.

And I wonder this specific advice to the audience of Def Con 2 builders in crypto how they can like differentially accelerate, how they can contribute like in a direct way to improve the world. Like yeah, curious for any And then one very practical answer is that crypto is just like a very high-value test bed for a cyber and info defense technologies in very obvious ways of which we're we're very aware. And I think helping people better solve those problems is is always been one of my hopes can actually be a a bridge to building those technologies that could then be a helpful to people in many other environments. Yeah, hopefully prediction markets can help us get better timelines on any risk or any upside and and help us with those questions. Yeah, the probability of a cliff will have And was was SPF a cliff?

Awesome. Yeah, yeah, like I was going to I was going to plug some some other ideas. So, you know, look, be find the field and technology that you think you are passionate about and know a lot about and you think will most impact the future. And in whatever you do, try to aim for good outcomes. At the end of the day, it's the people that are building things that end up writing a lot of how these technologies end up impacting billions of people in the world.

So, the values that you imbue into the technology that you build have a huge impact in the in the future. And so, today we're massive beneficiaries of the values of the people that built the internet and built the web and built so much of what we use day-to-day. So, as you go and accelerate some technology, imbue it with good values to end up with, you know, safe outcomes and the best benefit to as many beings as possible. Though, I do agree with the utilitarianism is probably not the right moral philosophy for the future. Uh but, you know, overall, like try to do good things um and try to like be conscious of where the bad parts might be and avoid those.

Awesome. That was a really nice ending. Thanks everyone for coming.