Solving Global Coordination
Date Posted:
September 29, 2015
Date Recorded:
September 29, 2015
Speaker(s):
Jaan Tallinn
All Captioned Videos CBMM Special Seminars
Description:
Jaan Tallinn is an Estonian computer scientist who participated in the development of Skype in 2002 and FastTrack/Kazaa, a file-sharing application, in 2000.
He graduated from the University of Tartu in 1996 with a BSc in Theoretical Physics with a thesis that involved traveling interstellar distances using warps in space-time.
Tallinn is a former member of the Estonian President's Academic Advisory Board. He is also one of the founders of the Centre for the Study of Existential Risk , and the Future of Life Institute , and was co-founder of the personalized medical research company MetaMed
PRESENTER: Let's start. This will be an unusual talk, I guess, because it's more of a question and answer and discussion than a real talk; no slides.
JAAN TALLINN: No slides.
PRESENTER: No slides. So it's great to have Jaan here. I've been asking him to tell me when he was going to be around. And he told me a few days ago. But I hope we can repeat that in the future. Anyway, he's one of these visionary people. By the way, he's, I think, 25 years younger than I am, so therefore, a visionary in the era of internet, one of the founders of Skype.
And Skype, it was about 14 years ago. It's interesting that the modern era of internet and machine learning, and so on, started when the internet bubble collapsed. This was 2000. And there was a lot of destroyed dreams at that time. But new ones came up, which have quite a lasting power. I mean, Skype is almost a word in the dictionary by now. You know, it's difficult to think of a time in which Skype did not exist. And here we are. And so it will be very interesting to hear from Jaan what he's thinking about, especially about collaboration, and--
JAAN TALLINN: Coordination.
PRESENTER: Coordination. So, Jaan?
JAAN TALLINN: Thank you. And good afternoon slash evening.
PRESENTER: By the way, you must be one of the few people who was born in Tallinn and is named Tallinn-- same name, right?
JAAN TALLINN: Yeah. All these people are my relatives. Yeah, I don't think there's anyone in the world who shares my name, full name. So one of the, I think, the most important things that I'm doing with my life these days, I've been doing in the last six, seven, eight years, depending how you count, is trying to save the world, literally. And so I've been supporting a network of people and organizations who are trying to lower a so-called existential risk, risks from technology, that if they materialized, would sort of put humanity in a very bad situation.
Now, there are two ways to address the existential risks. One is to do this whack-a-mole thing and look at each risk individually and then build capacities, which are addressing those risks. And in some ways, I've been trying to do that, trying to help people who are trying to do that.
Or you can basically take a step back and think about, how can we actually to make the world better, in general? In order to [INAUDIBLE] would be better in addressing those risks, you wouldn't have to have these crazy people, who get often put in the same pocket with doomsday, of doomsday prophets, and then, we have to defend ourselves. But the world actually would be-- it would be great-- wouldn't it be great if the world actually spent more time, more time and resources and talent, on making sure that humanity saves this-- humanity survives this century, that it spends on tobacco ads, for example, which is currently not the case?
So one of the projects that I've been on and off, doing in the last year or so, is putting together a document with ideas how to improve the world's ability to address various problems, in general. And although this document is still a fair way of being completed, I already have started extracting presentations from it. So this particular talk, I gave in Oxford, what, a month ago, at an event called Effective Altruism Global Summit, where young people from the movement call it Effective Altruism, which, I guess, many of you know, as you got around and try to figure out how to increase the amount of good that they do in the world.
But yeah, as Thomas I think mentioned, the talk is very short. It was meant to-- there, in Oxford, it was meant to be an introduction for a panel. But I hope it's provocative, I guess. As much as I have given it, it has always resulted in a lot of questions. And at the [INAUDIBLE] in Oxford, we could have had it going for a couple of days, I think, because there were so many questions and exciting ideas that came as a result of this topic.
So the talk-- "Everyone is hurting each other. The planet is rampant with injustices. Whole societies plunder groups of their own people. Mothers imprison sons. Children perish, while brothers war. What is the matter with that, if it's what you want to do? But nobody wants it. Everybody hates it. Oh, well, then stop."
So this dialogue originates from a quasi-religious text, called "Principia Discordia." But I lifted it from an essay called "Meditations on Moloch," by Scott Alexander. In that powerful essay, Scott points out that almost all humanity's ailments, problems like crime, corruption, overfishing, deforestation, global warming, arms races, and tropic anthropogenic existential risks, and so on and so forth, have a common theme the kernel. And that kernel is bad Nash equilibrium, a [INAUDIBLE] situation where people follow local incentives to a predictably bad overall outcome. So in some sense, such problems are all forms of prisoner's dilemma, where individual players in the game can't improve the outcome unilaterally. And they need external coordination to improve the outcome.
Now, an interesting question is that, is there a way to systematically crush this evil kernel that Scott, in his essay, calls the Moloch, and thus solve most of the problems that plague humanity? And I believe there might be. The key here is that none of those [INAUDIBLE] were deliberately put in place. Rather, they all emerged over time for random historical reasons. So no, there is no powerful institution or agency, in general, that has a vested interest in maintaining those equilibriums.
Or as my friend Michael [? Vasser ?] once put it, the world is like a race car speeding towards a cliff. But the good news is that there is nobody at the wheel.
[CHUCKLES]
Of course, we don't want to create a global dictatorship. Instead, we should engineer a sort of cruise control, that would automatically course correct the world whenever it deviates from pursuing human [INAUDIBLE]. Of course, this sounds very hard. It is far from hopeless. There are numerous historical examples, for example, of societies handling tragedy of commons situations by culture or peer pressure, or sometimes even using clever, self-enforcing protocols.
So for instance, in Athens, in ancient Greece, they had a system called liturgies, where the richest people had to finance public goods. Now, in case they refused by saying that I'm not the richest, they had to agree to exchange all their possessions with a person who they thought should do the financing in their stead.
[CHUCKLES]
Well, such clever solutions to game theoretic problems are not only getting easier with new technology-- think Wikipedia and blockchain-- but also, organizations such as Global Forest Watch, who crowd monitor deforestation in real time, using satellites and the internet. One way of looking at Wikipedia is that, from a game theoretic perspective, what it brought on the table was that it flipped the cost of vandalism, at least simple vandalism, versus reversing, or fixing, vandalism, thereby allowing knowledge or structure information to accumulate without it being vandalized.
But in general, the trends in technology point towards a coming explosion in transparency, which brings new chances, of course. Yet, it clearly simplifies accountability and enforcement. Last, but not least, the effective altruist movement, that I gave the talk to in Oxford, the movement itself carries the potential to change the game theoretic situation that the world is in. Because after all, effective altruists define themselves as people who prefer global outcomes to [INAUDIBLE] the local incentive gradients.
So in the bird's eye view, what I'm proposing is to combine new technology with humanity's past experience solving various domain-specific coordination problems in societies and various domains, and engineer a general solution to the world's problems. Such a solution, or more likely, set of solutions, must have, in my view, at least two big components. First, there has to be a preference discovery mechanism for figuring out what qualifies as a good future, according to human values, in the first place, and second, a preference promotion mechanism that would incentivize people to contribute towards such a preferred future. And I do note that the problem is similar to developing a value aligned AI.
First, figure out what you must want, and then build a system that optimizes towards those values. Having the preference discovery mechanism separate from the promotion mechanism would serve several purposes. First, people are rightly suspicious of movements that just strive for power while confidently claiming to know what the correct values are. Second, doing value discovery in a transparent and trustworthy manner might create the sort of a general metric, external metric, to gauge the actions of prominent people and organizations against.
And third, aggregating human values would likely help to solve the ultimate AI value alignment problem, not to mention that they actually do expect pretty good on our progress initially. Because I don't think much effort has gone into aggregating human preferences. So a new margin of work in this area should yield pretty nice results.
So what global mechanisms might have the effect of discovering and promoting human values? I don't know exactly, of course. But in my research and thinking over the last year, low intensity thinking even, I have come across already several interesting ideas. For instance, Nick Bostrom, who leads the [INAUDIBLE] in Oxford [INAUDIBLE] suggested that we should perhaps try to create the cultural tradition of rewarding past actions, actions whose importance might only be revealed later with the benefit of hindsight.
For example, just on Saturday, with a few people here, we celebrated what's called Petrov Day, to memorize the-- or to remember the events of September 26, 1983, where Stanislav Petrov, an officer in Soviet army, decided not to launch a nuclear counterattack after, what turned out to be, a false alarm of Soviet early detection system. Many of you, and certainly me, wouldn't be here if he had decided otherwise. So in some ways, sometimes I think that it's weird that my existence has-- the cause or path to my existence, continued existence, has gone through the brain processes of a single man.
So yeah, the thing is that what Nick Bostrom suggested is that it's likely that the future societies are going to be richer than the current ones-- not to say anything about the distribution of the wealth, which we might have a problem with. But on average, they will be richer. So resources shouldn't be a problem. If you create the expectation that, if you do something good for the future generations or future society, you would get rewarded. That actually should push the world already into a better match equilibrium.
Now, whether or not our idea is Paul Christiano's certificates of impact, they are basically-- the idea is that, whenever you do something good, that I think that might be good, that might be appreciated in future, you can claim a certificate for having done that, which you can then sell to potential donors later. So the mechanism helps people to do good deeds now and then raise money for it later, once the effect is clear. And I think that such certificates would actually be especially powerful if there was a futures market for them, plus a set of retro charities that specialize and commit to buying the certificates in proportion of their future impact, of their eventual future impact.
And finally, I have an intuition that combining blockchain technology and decentralized prediction markets with global mobile e-voting could yield the potential preference for [INAUDIBLE] mechanism. The idea is that, for example, we know that people are not very good at knowing what they want from the future. And things go completely haywire when you ask them, what policies should we enact in order to reach a good future? Because then, it just becomes a signalling game. People are just signalling what kind of people I'm allied with in this pick a policy game.
But however, what if we created a polling system that's global, that globally provided in a randomized manner, polls people's current well-being, like asks them to ask-- asks them to respond, like how well are things going, in general? And then, if they aggregate that [INAUDIBLE] number in some magical way, but also a transparent way, using some kind of decentralized mechanism, and [INAUDIBLE] the futures market, that would predict the effect of different policies on that particular number, condition on those policies being enacted. So we could actually have people putting their money on the line in order to predict the effect of different policies on this global welfare.
So in summary, as Scott Alexander wrote in his "Meditations On Moloch" essay, many of humanity's ailments originate from accidental gain [INAUDIBLE] equilibriums. Yet, humanity has never been completely helpless against such equilibriums. We have a history of defeating and fixing them in many societies and domains. It's clear to me, though, that in this century, such retroactive domain-specific batching is just no longer sufficient. We now need global mechanisms for humanity's preference, discovery, and promotion.
One thing I really like about effective altruists, in particular, is that they are always hungry for opportunities to maximize their positive impact on the world. So that's why so many of them actually end up focusing on existential risks, because they think that, well, one way of making my good impact on the world or improve my impact on the world, impact on the world, is to go from local to geographically diverse. So instead of just helping my local community, think about what can I do with my dollars or time in the world where those are most needed, and thereby increasing?
And many of those effective altruists are actually starting, wait a minute. Once I'm starting to diversify [INAUDIBLE] no local in space, shouldn't I also go non-local in time? I think about what would the future generations want me to do? And hence, many of them actually end up interested in lowering existential risks, because the assumption is that-- actually, that's the punch line here. I'm not going to say it? Helping to develop humanity's preference, discovery, and promotion mechanisms has the potential to have even greater impact than existential risk reduction, because, assuming humanity values its own survival, it addresses actually [INAUDIBLE] reduction just as a special case. So let's do it. Let's dethrone the Moloch.
[APPLAUSE]
AUDIENCE: So what would be the first action item?
JAAN TALLINN: Yeah, so my own plan here is to-- because over the last few years, I've collected-- I've been connected to a set of organizations and people that I support in many ways. So my current plan is to just finish the document that I'm writing and then just solicit feedback from them. And the main idea of this document is not going to [INAUDIBLE] prescriptions about what we should do. It's just meant to be a collection of ideas and just literature references, to point out that, look, there is something there. Let's go and look for it.
And once I do get the [INAUDIBLE] feedback about this document, I hope that we could divert some of the research that Future of Life Institute, for example, is funding, towards more general research, just not being focused on AI safety or bio-safety. Just let's just take this thing one more, one level at a time, and do some research internally, kind of improving the work. Because there really seems to be something there. Let's figure out. But yeah, so these are the first two steps that I'm having in mind.
AUDIENCE: Is it [INAUDIBLE] you really [INAUDIBLE] about the technology and [INAUDIBLE]
JAAN TALLINN: Yeah, so I mean, like I said, all these are a bunch of existing research about how different cultures have handled so-called common pool resource problem. For example, how did they handle the situation where you had common grazing grounds, and what were the empirical rules that seemed to work, when it came to managing this common pool resource? So I absolutely expect that many of the interventions can be just purely cultural.
For example, what Nick Bostrom suggested is just-- this is just a cultural or mimetic intervention that doesn't have to do much with technology. You popularize the idea that instead of donating to charities that are currently doing good, you try to figure out who do we want the reward from the past? So this is just purely cultural intervention. However, having said that, I do think that technology is going to play a special role, because it provides a medium for doing a global scale coordination, a global scale-- you're enforcing global scale policies.
That's why I'm really interested in blockchain, blockchain technologies. One way of putting what blockchain really is, it that, in the last five years or so, we have lived in a world where it's possible to do global coordination or reach global consensus in a decentralized manner. In bitcoin, it's just basically a look-up table, like a key number, a key number, key number, that says who has how much.
But there is already a very nice generalization of that out there, launched a month or two ago, called Ethereum, that maintains a global state of key and just key value, no longer key number, but key value of pairs. And those values can also be scripts programs that can run and actually change the state. So you have this basic-- in principle, you have a [INAUDIBLE] complete, shared global state machine, which is potentially very interesting, because you can do commitments and contracts and hopefully, something that will help with pushing the global [INAUDIBLE] equilibriums into something more beneficial than they are now.
AUDIENCE: You also mentioned the idea in the global marketplace. [INAUDIBLE] Did you think about what mechanism would be to assess what those ideas need to do [INAUDIBLE]? Because OK, you've got an idea to do something, and [INAUDIBLE] you can sell it on the global marketplace, then you might have means to implement it. What would be the mechanism to assess effectiveness that you [INAUDIBLE]
JAAN TALLINN: Yeah. I have final thoughts there. But in some ways, I'm a part-time agent investor, right? So I see these situations in a commercial context. Some good teams come and say that, OK, we have this particular idea. And then you have to assess the quality of the idea, plus quality of the people. And if you find it sufficiently good, then you will basically give them money for a share in this enterprise.
Now, one idea would be you're going to internalize this mechanism that only works in an economic context, internalize that you're going to having a positive impact on the world context. So if we had this, what I call, retro, retro [INAUDIBLE], that would pay, in retrospect, to people on teams for having had a good impact, it became suddenly commercially interesting to look for teams that want to, basically, compete for the future rewards from those charities. And then, you basically have all this existing mechanism of assessing the quality of teams and ideas, to figure out how successful they would be. So that would be one of the ideas for assessing ideas in.
AUDIENCE: So I have a comment that I [INAUDIBLE] question [INAUDIBLE]. When Donald Trump comes out and says, this new tax plan of mine will not increase the budget deficit, and then some other politician says, yes it will, you can be sure that they're just going to keep saying-- sticking to those statements over the next year, no matter what happens. And however, you never see these kind of deadlocked questions when people argue about what the oil price is going to be in 30 years, because you have a very functional futures market.
And if someone is really convinced that oil is going to be much more expensive in 30 years, then the rest of the market thinks they're going to go buy and make a killing on it, right? So [INAUDIBLE] love the things you're saying about actually having people put their money where their mouth is and say, if they believe that this policy x is going to have effect y, there should be a market where--
JAAN TALLINN: Betting is tax on bullshit, as they say.
AUDIENCE: Say again.
JAAN TALLINN: Betting is a tax on bullshit, as they say.
AUDIENCE: A tax on bullshit, exactly. So my question related to this is, can you just summarize for us a little bit what the legal impediments are for having few betting markets and future markets? Because it's [INAUDIBLE] pretty strong [INAUDIBLE].
JAAN TALLINN: Yes. So futures or prediction markets, I think is the most general name, they have basically two big constraints to problems, that there is an inherent problem, intrinsic problem, is liquidity. If you don't get liquidity, then you don't have a market. And interestingly, Robin Hanson, who is a friend and economist at George Mason University, and he is a world expert on prediction markets, he has found out that dumb money is also OK. If you get dumb money, that actually gives you liquidity, and that's sufficient. You don't have to have smart money. If you have dumb money, basically, what it does, it attracts smart money.
So sometimes, even, he even proposes to get over-- if you really want the answers from a prediction market, and you don't have liquidity, it might actually be worthwhile to subsidize it. So you actually have a charitable organization or someone, who was a bunch of money, just put up-- insert with a bunch of money to kickstart the prediction market. So liquidity is one big problem that needs to be overcome in order to make the prediction market effective.
And indeed, the other big problem is legislation. I think Intrade was the prediction market here in the US that was just shut down, because it ran afoul of the gambling laws.
AUDIENCE: Oh, they've shut it down now?
JAAN TALLINN: I think so, in US. But that's my understanding. I haven't actually checked. So now, this is where the decentralized thing helps. I mean, right now, there is a decentralized prediction market called Augur, just being developed. They hope to launch next year. We'll see what happens.
AUDIENCE: Augur?
JAAN TALLINN: Augur, A, U-- I have to translate the letters-- G--
AUDIENCE: U-R.
JAAN TALLINN: A-U-G-U-R, yes. And that idea is that they are-- because a theorem platform, as I mentioned, is a blockchain technology that's true and complete. You can build the, as they call them, decentralized applications on top. So the idea with' Augur is that they're going to have a decentralized prediction market on top of a theorem blockchain. Therefore, there is no single point to attack legally or with--
AUDIENCE: You mean through the-- Tommy and I have a bet against each other about Donald Trump's tax plain, it's really just a private bet between us, and nobody else knows about it?
JAAN TALLINN: No, people will know. But no-- well, I mean, [INAUDIBLE] see it [INAUDIBLE] autonomous, [INAUDIBLE] was a bet. So people will know that the person with a certain key has a bet against a certain person with another key. But that's--
AUDIENCE: Then are we allowed to do that, Tommy and I?
JAAN TALLINN: I mean, that depends on legislation, depending on what country--
AUDIENCE: [INAUDIBLE] decentralized, so it cannot be shut down.
[LAUGHTER]
JAAN TALLINN: Yeah, just like a [INAUDIBLE] business idea behind Kazaa was that-- Kazaa was thought of when Napster was in the process of being shut down. And it turned out to be much more harder to shut down Kazaa. It was peer to peer. Then eventually, they still managed to do that. So yeah, well, in general, I think the decentralized currencies, decentralized prediction markets, they are not like a panacea. So they are just new coordination mechanisms for humanity that have different properties, different trust, and different, in general, different nature than the existing coordination mechanisms.
For example, UN-- it is sort of a coordination mechanism. It has been good in very few things, but it has managed to do a few coordination feats-- ozone, addressing those, and coordinating the CFC reduction, you know, to say those. And [INAUDIBLE] I think, was one of the triumphs of UN. And also, the decentralized prediction markets, they come with their own set of failure modes. For example, assassination markets is one known failure mode, where you basically have this anonymous way to bet on people's expected lifetime, which is terrible, because you incentivize people, insiders, to bet on a certain date, and then have-- kill the person.
So yeah, it is just a tool. We need to figure out how to avoid a lot of bad, failing [INAUDIBLE] and those [INAUDIBLE] tools, as well. But they also are a good opportunity to build something better.
AUDIENCE: The markets like this should, in principle, make elections, say, political decisions, easier, but so far, has not led to a [INAUDIBLE] there was [INAUDIBLE] the political prediction [INAUDIBLE], which, by the way, had an exception allowed by the whatever federal regulator.
JAAN TALLINN: Where was that?
AUDIENCE: This was for predicting your betting money on, say, it was Bush versus Clinton, for example.
AUDIENCE: Yeah, that was Intrade [INAUDIBLE], that got shut down [INAUDIBLE].
AUDIENCE: This was earlier.
AUDIENCE: Oh, okay.
AUDIENCE: And this was an academic problem.
JAAN TALLINN: Oh, OK.
AUDIENCE: And so they got the permission to do it. They did it for at least two or three elections. And usually, the results are OK. So the money involved were like $100 or [INAUDIBLE]. And unlike elections, this, over time, you see what other people are betting on. So there's much more information [INAUDIBLE]. But I don't think this convinced [INAUDIBLE] that's the way to go.
JAAN TALLINN: Yes. So one-- sometimes, I joke about Robin Hanson, who is expert on prediction markets. These are professional physicists who turned into professional economists who turned into professional cynicists. Because he exactly-- basically, his progression was that the he had a interesting toolset in physics. He thought that he can bring it into economics, economy, and do various experiments and build up various tools that companies can use to make their operations more efficient.
And then he did that and found that, hmm, companies are not really interested in that. [INAUDIBLE] But hmm, what's really going on? And then, all the other interesting work actually happened that he has been doing. So for example, with prediction markets, he says that, within companies, especially big companies, the big problem in prediction markets is that they are second-guessing the leaders. And leaders, it would be really bad for leaders' status to have basic there [INAUDIBLE] caught. So there is no-- on top of big companies, there's just no interest in prediction markets for that simple, social, tribal reason.
So yeah, we actually have to be aware of weird human psychology things that might be an obstacle. But also, it might be an opportunity, in some cases, where something goes viral for some reason, et cetera. So yeah, you have to be smart about prediction markets and other coordination technologies.
AUDIENCE: [INAUDIBLE] with what you say. To use a prediction market, I mean, that would be a clean purpose there.
JAAN TALLINN: In what [INAUDIBLE]?
AUDIENCE: It is a little problem in using the prediction marketplace for purposes that you [INAUDIBLE]
AUDIENCE: [INAUDIBLE] financial purpose. You're trying to do something good. [INAUDIBLE] something good not necessarily has a direct financial [INAUDIBLE]. If you could [INAUDIBLE] money with it, yes. So it is more like a [INAUDIBLE]. For example, if you paid money to people who [INAUDIBLE], you have money to incentivize.
JAAN TALLINN: Yeah, yeah. So I mean, actually, just yesterday, it was yesterday, where we looked at what is the US budget, if you look at the discretionary budget that the government can spend on various issues? And then, on the other hand, when you look at the amount of money that people give to charities per year, the latter is actually much bigger. So the idea there is to-- to [INAUDIBLE] the problem that you bring out is to figure out how to convert a bunch of these donations to charity, to [INAUDIBLE] that actually incentivize fuels. Things like prediction markets are actually more-- basically other mechanisms of doing good better than the current charity mechanisms some charitable organizations are doing. So that's one idea, at least, to tap into the charity money.
AUDIENCE: So what if this is market-based [INAUDIBLE]? Would society be a kind of threat that's very unequally distributed? Doesn't this mean that you have the same problem of being [INAUDIBLE] now that money was [INAUDIBLE], so that you can-- I mean, you can rig the game the same, the market in the same way [INAUDIBLE] put other money much more influence than [INAUDIBLE]. And you can influence the way, where the money goes [INAUDIBLE]. So how does this fix [INAUDIBLE]?
If I understand you correctly, you're saying that use the prediction markets to fund charities or fund good causes. So in the end, if someone has a [INAUDIBLE] interest that money goes to supplement [INAUDIBLE] direction, is there [INAUDIBLE] the possibility [INAUDIBLE] market that he has much more [INAUDIBLE] How is this [INAUDIBLE] know that [INAUDIBLE]
JAAN TALLINN: So there are actually multiple questions in that, now. One thing is that, indeed, it's [INAUDIBLE] it's possible to rig prediction markets, if you are willing to spend the money. Basically, if you really want to enact certain policy, for example, just let's say that there are policy A and policy B, and there's a choice between those. And a prediction market basically says that if you enact A, is the outcome good, or is it going to be a failure? And if you enact policy B, is the outcome going to be good or bad?
And so, in a neutral market, you basically get [INAUDIBLE] an aggregated, expert answer, just like as you get in, as Max mentioned, the futures market for oil, for example. Now, if there's very really strong vested interest in enacting policy, then we also know what happens. We know what from the stock markets. It's called cornering the market. But when you're going to throw a lot of money at the problem, you basically can bias the market, the prediction market. However, the problem there is that you're going to lose that money.
So you biased option A in order for it to be enacted. But then, it goes on to fail, because that would have been a [INAUDIBLE] prediction, unless there's some weird self-enforcing thing there. And then, you basically are going to lose the money. The outcome is going to force you to pay out the money to the people who were betting against it. So it's a known failure mode, and we need to think about it.
Actually, the worst, the even worst failure mode of prediction markets-- and almost, I think the general point here is that the prediction market's just like one tool in the set. I'm don't want to really get really hung up on those. There are many, many other ideas how we could create a better corporation and a better value, value discovery mechanism in the world. But prediction markets there, they are fascinating. So what was I going to say?
So yeah, the one failure mode of prediction markets that I think we should figure out how to address is that they are not very good at predicting existential risks. So if there's a certain probability of the world blowing up, you should always, at least in a [INAUDIBLE] prediction market, you should always predict against it. Because if you predict for it, there will be nobody to collect your winnings from. So there is this issue that we should figure out how to address.
And then the other part of your question is how can we ensure that this-- I guess the way I would rephrase your question, how can we make sure that the value, what I call preference discovery mechanism, is fair? It's not biased towards a white male in Silicon Valley, right? And it's a non-trivial question. But I think what would really help is that-- the nice thing about mechanisms like blockchain is that they're very transparent. So if you can tap into that transparency, then there might be something there that would give a perhaps, even almost [? probably, ?] a nonbiased value aggregation; but more research needed.
AUDIENCE: [INAUDIBLE] it seems like a lot of the solutions they propose are some kind of technological access. If we're not able to solve in an economic inequality, then how are we going to have widespread participation?
JAAN TALLINN: Cell phones.
AUDIENCE: What's that?
JAAN TALLINN: Cell phones. Cell phones are basically a huge explosion in Africa, for example. And so they're going to-- yeah, I do think that internet access is going to be very ubiquitous, especially now with satellites going up. And basically, we're going to have world-wide Wi-Fi, or something like that.
So I'm pretty optimistic about people getting connected at an increasing rate. However, I might be wrong, in which case, we need to figure out some other mechanisms. The key thing is that, for value, preference discovery part, you don't actually need to poll everyone. It's just your sample has to be random, random enough. And if it's random enough, and yet, the number of people is small, you can just travel to the villages and ask them.
AUDIENCE: You have been speaking mainly from the point of view of individuals getting together and cooperating or funding enterprise. What about getting governments involved? Is there hope of committing one government to something interesting to set an example? [INAUDIBLE] what about Estonia?
JAAN TALLINN: Yeah, yeah.
AUDIENCE: Is the most digitally connected?
JAAN TALLINN: A couple of years ago, I actually imported Robin Hanson to Estonia. So I introduced him to a bunch of Estonia officials, in the hope that they would pick some ideas from-- so far, no luck. But Estonia, indeed, is in an interesting position. Because what's going on in Estonia-- first of all, it's a small place, so there are many politicians. Most--
AUDIENCE: What is it, 1.5 million?
JAAN TALLINN: 1.3, and the country is the size of Manhattan, when it comes to actually a population. So the nice thing about this is that there isn't a lot of selective pressure when it comes to getting political officers. So you don't have to be raised as a politician from the womb. You can actually-- most politicians have normal jobs, as well. So they're going to bring their experience from different fields. And then, they say, OK, what can I do with my toolset from my previous profession in non-politics?
So that basic context has put Estonia in the place where people have started experimenting with e-governance. And for example, I voted with my mobile phone for the last, I don't know, six to eight years. And so there is a sort of culture of innovation in governance circles that has gotten them a lot of positive feedback. So Estonian officials are being asked to present at conferences, and they have been asked to advise the UK government, et cetera, et cetera. And when they do that, then they come back to Estonia, OK, that went really well; let's do something else. So [INAUDIBLE] in a nice, positive feedback.
The latest thing is that now, they have this program called the residency. So basically, you can go to-- I don't know if all, but to some Estonian consulates, and get yourself this ID card that you can use to give digital signatures that have the power of law in Estonia. And therefore, you can create your businesses. And they're still working at banks. Banks are a little bit reluctant to have people open accounts without showing up. But I think they will, at least at some point, relent.
So yeah, I do think that places like Estonia or other forward-looking governments might be very helpful in figuring out how can we do global coordination? That said, as my friend Patrick Friedman, who had this free cities movement, he said that the US Constitution is from 17th century, I think. And if we drove a car from that time, we would be a horse. So that the principles of government are really, really, really old. They're not really designed for the current time. So most of the mechanisms from there are just worthless, for global coordination. But some of them might be useful, and we might want to scavenge those.
AUDIENCE: I mean, if you look at laws from the point of view of software engineering, say, the tax law in United States, to be specific. After fixing a few [INAUDIBLE] over the years, at some point, you should fall to the cause that [INAUDIBLE].
JAAN TALLINN: Yeah, yeah. So actually, one nice thing about Estonia is that we got a fresh start in '90s, when internet was already around. So we didn't have all this legacy. So I got the first [INAUDIBLE]. Now, we have basically all the-- start seeing some signs of the accumulated legislation. But that is a huge problem in US, I think.
AUDIENCE: So as a software engineer, what should be [INAUDIBLE]? The software have [INAUDIBLE]
[LAUGHTER]
JAAN TALLINN: Well, that's really interesting. We can now talk about--
AUDIENCE: [INAUDIBLE]
AUDIENCE: There is an explanation time, when you think about it.
JAAN TALLINN: Yeah, it's all very interesting questions [INAUDIBLE].
AUDIENCE: [INAUDIBLE] maybe constitutions.
JAAN TALLINN: Yeah, you want to have some kind of being able to react. I mean, it's a huge problem in any way. It think Bruce Schneier, a few years ago, gave a talk at some conference that I attended, where [INAUDIBLE] he pointed out that legislation and enforcement is always playing catch-up game now, because they're just much lower than all the criminals and hacking at the age of information. So it's a general problem.
AUDIENCE: Yeah, I'm wondering, there are countries like Switzerland, who are doing [INAUDIBLE] actually, I'm not sure there are many other examples of doing what you want economically for quite some time. And they have constant referendums. I don't know whether there are other examples [INAUDIBLE]
JAAN TALLINN: There are some interesting governance innovations that have tried. For example, in my document, one of the things that I mention is liquid democracy. It's a very cool idea. I'm not sure how well it works, but it's still worth investigating. And I think they use in Germany and some, I think, [INAUDIBLE] party there, or somewhere. It's basically the idea that, instead of having the regular elections, everyone has one vote, and he or she can just give that one vote to someone else, to anyone really. And anyone with a collection of votes can just again distribute these votes to other people. And then basically, those people who end up having the largest collection of votes, they are in power.
However, the thing is that, whatever they do, people can always pull back their votes that they gave them. So So there's a dynamic system of like real time accountability, which is-- and the other really cool ideas are just randomized. There is this-- if you know the fundamental problem in representative democracy, like the [INAUDIBLE] there are three, three reasonable properties that you want an election system to have. And they can show that they if you're a representative democracy, you're going to violate at least one of them, which is bad.
However, once you introduce a randomized element, once you introduce an element that voting is not deterministic, there is a certain element of randomness you can get around of those and actually, in some ways, get a much better representative democracy. So I think there's a lot of--
AUDIENCE: And there are a lot interesting issues. If you have electronic elections or [INAUDIBLE], then, of course, the first thing to do is [INAUDIBLE] checks [INAUDIBLE] is a person not at home. But the next thing would be, is the person who is voting, did he read what he's voting about? Does he know [INAUDIBLE] not an intelligence test, but a knowledge test?
JAAN TALLINN: Yeah, so that's why--
AUDIENCE: I wonder how many people passed it?
JAAN TALLINN: That's why I think it's--
AUDIENCE: [INAUDIBLE]
That's why I said, [INAUDIBLE] was that [INAUDIBLE] electronic voting system, you can always-- before the elections are committed, you can always go and change your vote. So even if you're forced to vote for someone, you can basically go and then have a paper ballot to change your vote. And then you can check, I think-- I never checked, but I think you can check, what is the status of your vote right now? [INAUDIBLE] I don't think Estonia is using that, but there are interesting cryptographic ways to actually prove that your vote hasn't been tampered with; so again, ideas to use there.
However, the general answer to your question is that-- I think, as I mentioned in my talk, we really want to actually disentangle the preference discovery mechanism and policy discovery mechanism. So like, when people know what's good for them right now, doesn't mean that they know what's good for them in the future. And even worse, most of the people have no idea what policies. So it would be great to actually do the policy search separately from the preference discovery.
AUDIENCE: [INAUDIBLE] politician in this country obviously no taxes or some taxes. This would be a modern running in showing what are we after.
JAAN TALLINN: Yes, yes. We basically want to create hedge funds in the policy space. And that will make predictions about what are things. Because I think there's so much talent being wasted in which hedge funds right now. I think the zero sum gains that, if you could just get a fraction of those working on actually what matters in the future. I've done my own [INAUDIBLE] trading, at some point, so I'm partly at fault.
PRESENTER: OK. You have a flight to catch. So let's thank you.
[APPLAUSE]
have an interactive transcript feature enabled, which appears below the video when playing. Viewers can search for keywords in the video or click on any word in the transcript to jump to that point in the video. When searching, a dark bar with white vertical lines appears below the video frame. Each white line is an occurrence of the searched term and can be clicked on to jump to that spot in the video.
