Managing Digital Property Rights using Blockchain - Co-Pierre Georg

Transcript

[Music] yeah welcome everybody thanks for being here I know it must have been a long day um I'm going to try to make it worth your while by talking a little bit about uh the foundations of data and the role that blockchain can play and thinking about different data infrastructures and I want to start with two observations with two trade-offs that we have accepted that I think we have accepted exactly the wrong way around and this is really sort of the culmination of quite a lot of research um over the past couple of years of myself and my working group and it's actually part of a book that is going to come out later this year together with Katarina pisto at Colombia so a lot of the things that I say if they're very

clever they are her ideas and if you think they're wrong they're all mine so just keep that in mind I want to start with these trade-offs the first trade-off that we have accepted is that our data is available pretty much to anybody including you know commercial Enterprises but their code remains proprietary by and large we have no idea how the search algorithms precisely use our data we don't know how social media precisely ranks news feeds we have some idea there's some research on it but we don't actually know specifically how our data is used so the data is available to everybody and the code is AA better okay so the data is available to

to everybody Who can capture it and the code is available only to those who have control over the data I think this this trade-off is exactly the the wrong way around it should be the other way around our data should be available only to a very select few and they should be able to do with it only things that we approve of that we know how they do how they handle our data the second tradeoff is related to that the trade-off is that we've accepted that access to data is narrow that you know few people have access to data think about gdpr when you've worked with any kind of research data you will have encountered that issue that the access to the data is narrow is restricted to a few people and it's restricted to a few people because once you have access you can pretty much do whatever you want

with it there are some constraints some technical some legal constraints but both are difficult to enforce given the existing information infrastru structure and again I think the tradeoff that we have accepted is exactly the wrong way around access to data should be brought to many people who should then be able to do only the things that those who have created the data actually agree with so that access should be Brad but you should retain control over your data so these are two two trade-offs that we have accepted and I think with new technology and that's sort of the core of the talk today is to discuss how sort of the information INF structure that we have can be thought of differently using

new technology that we have with new technology we can actually think a couple of things differently from the way we have implemented them historically and it it begins to matter when we first started building information infrastructures moving you know from the mainframe computers um to personal computers to the cloud data didn't have that kind of value that it has today the first sort of main frames when you go back to like even the very early days of the first computers anak edak and so on you know data didn't really have value it was there sure and you would do some things with it sure but the first time that data only became valuable was when we did realtime

Computing with Airlines um and in um in some banking applications so only then was Data valuable so the infrastructure that we have today and I'm going to talk a bit more about how that looks like and what the components are this infrastructure was built during a time when data wasn't actually that valuable especially not relatively uh relative to the infrastructure that would process it so the thing that had value was the expensive large you know room filling kind of Machinery that would process the data but not the data itself nowadays it's exactly the other way around nowadays data is much more valuable because we have found new ways to use it so the infrastructure that we

have to manage to maintain to process to store the data was built during a time when the thing that it would store wasn't actually that valuable so there is a disconnect here and this with AI I think it's it's really coming to a point where um I had this talk the other day where he was where he was speaking about how it's not about an arms race about chips it is an arms race about high quality data and there sort of when you when you follow the Deep seek um discussion over the past couple of days it's the first glimpse of it's not actually an arms race about who gets more Nvidia chips it is an arms race about who creates a data ecosystem that creates high quality good data that you can use to train models

because then you can substitute the chips and you don't actually need to spend hundreds of millions of dollars for training a model but maybe much less so if we have if we enter this world where you know bad data really becomes an issue I think it's worthwhile to to to rethink how do we process data how do we store it how do we handle uh data how do we build information infrastructures and I want to give you this one sort of like high level picture of this the fundamental problem that we have both from a computer I'm an economist by training most of the time anyways um so both from the computer science perspective as well as from the

perspective of of people think think about incentive systems is that if you have data that is controlled by firm a and you have data that is controlled by firm B and you want to bring it together you are screwed because either firm a needs to share their data with firm B so that firm B can do with it whatever they want remember this trade of that we've accepted or the other way around or both of them need to share their data with a trusted third party there isn't a lot we can do other than that I'm going to talk later a little bit about like some new cryptography methods and like in CS there's some research on how to deal with it I have some thoughts I'm going to share about this but it's it's really hard to bridge this fundamental problem that whenever you copy data whoever gets

a copy of the data can basically do whatever they want with it partially because you know legal Frameworks are weak to are weak and are difficult to enforce because the action of you know copying data illicitly can be unobserved so that's the that's the fundamental problem that we have that you that we need a new kind of technology to actually combine this data what I'm going to argue later is that that's not enough so crypto has figured out some ways to do this fully homomorphic encryption Federated learning there are some things we can do to actually do this combination that alone is not enough because as long as you have operators of this infrastructure you

always have the problem that whoever controls the infrastructure controls the data that is processed in it and that's where the system design that I'm going to present towards the end of the talk will actually try to deviate and say well maybe there's a way that we can think about it differently and blockchain does play a crucial role there and this this this problem really matters I I worked at the German Central Bank uh for almost 10 years um we had the issue that we were collecting as a central bank data in different divisions so there was a statistics division a banking supervision um banking supervision Department both of them were collecting data and we wanted to merge it for financial stability analysis doesn't seem like such a crazy thing

because it's the same institution the same legal framework it's The Regulators you know we should be able to do this turned out we're not because both divisions were collecting data in their different legal uh framework and as a consequence we weren't allowed to match data collected about the same Banks from divisions internally the result was like a 2year internal battle over how do we deal with that and the the solution was we create a new division with a new mandate that would match data the research data service center so that's fine if you are a center bank and you have infinitely Deep Pockets and you can do things like that but it's not a good solution or slow it's cumbersome and

it's a reflection of this of these trade-offs that we have accepted so because you can't really merge data because it Still Remains siloed it it impedes your research left right and Center we've we've done some uh some pilots during covid uh with medical researchers about privacy preserving contact tracing they had fantastic medical trial data they wanted to bring it together with other data they just couldn't why because privacy regulations would prohibit it so the infrastructure that we have has created a legal framework that makes it expensive and cumbersome to give us a very shallow sense of protection and control over our

data and that is exactly sort of the criticism that we are leveling against this system but it's sort of not enough to to level this criticism I think we can do a little better and that's exactly where combination of this new tech can actually help us so by the end of this talk I hope I can sort of leave you with five sort of provocations if you want right think that that might not be easy to swallow but that I nonetheless think are true the first is that privacy is a fundamentally bad concept it's badly badly conceptualized both in legal terms as well as in technological terms by accepting privacy as the premise for data

control we have in essence accepted these tradeoffs and we think that's that's wrong we shouldn't accept that privacy is not a concept suitable for data or information the second hypothesis is that the open source movement failed they managed to protect the openness of the code but they failed to protect the data they prevented Microsoft from enclosing all the code fine but they didn't manage to prevent them from enclosing all the data and Microsoft was just fine with that trade-off and other big tech companies were just fine with that trade-off the third provocation is that the web the web itself is fundamentally flawed

because of the way we have implemented hyper text as one directional linking there is a history and when you go back to and that was sort of the the joy of writing that I just got tenure took sabatical we decided to write a book you know how as you do the joy of doing this that I got to go through like some of the really old debates in history of technology and and and read about like what were the Alternatives back when hypertext was first developed there were viable Alternatives proposed by folks like Ted Nelson who said well if we if we have one directional linking we will always lose provence of information we will never be able to fully capture the nature of data so the idea that Tim bernes Lee in

essence s of pushed with hypertext as one directional linking I think is a really bad one and unnecessarily so and there's Alternatives all right the fourth one is maybe a little less sort of controversial it's the technology constraints law and economics I'm actually going to start with this as a premise for everything else I'm going to do technology is not neutral or maybe this controversial in this C I don't know Libertarians don't tend to like this if you say technology is not neutral um but I'm I'm going to make an argument for this and the last one is that you know if you think about how to fix this it helps to go back through the history of Technology Point your finger at the things that have gone wrong look

at what were the Alternatives proposed at the time you know to not go completely crazy and only think about things that were actually feasible and possible and then just take it to the next stage and and come up with an alternative alternative system and I think blockchain does have an important role to play in this fix in fact I would make the argument that it is the only useful application of blockchain and that I have haven't met any other that I can defend after working on it for many years both as a regulator as an entrepreneur as a professor so yeah happy to happy to also take you know push back this is not supposed to be know me just talking about this and you just yeah yeah and

then walking out and disagreeing so please feel free to S of interrupt disagree at any point all right so that's the that's the plan for for my talk and I think I have like 30 minutes or so left all right so let me do let me start with the first argument just just a premise of everything I'm going to talk about and the reason why an economist should even be allowed to talk about technology you know which is not self-evident and that is that technology itself is not neutral langon winner in the whale and the reactor he makes that argument saying that technology shapes Society there's lots of evidence that this happened think about uh the recent paper about how Facebook causes hatred

against migrants so they use a quasi natural setup where there's random outages of Facebook to show that this predicts um anti-migrant violence in Germany so we know that technology shapes Society we also know that Society shapes technology think about the penny far thing the you know H in in German yeah the bicycle that women couldn't ride because their dresses kept getting stuck in the in the big wheel was it it's sort of in the history of technology it's like this ex like the standard example why technology is not neutral why the society and the norms and sort of everything we have in society shapes the way technology is

created libraries are another example Li and and Noam Chomsky makes this argument he says you know libraries are basically a neutral technology because anybody can go in and find whatever they want ironically his own books were banned from libraries across the US military basis for a long time because he said things that the US military didn't like so he more than anybody else should know that libraries are not neutral technologies that they curate information rather than just presented in the decision what you put on which shelf what you group together how many books you even include in your library you take a decision you influence what

information is presented in the same way that search engines do that so technology I think is never neutral and we must accept this as the premise so if it's not neutral then there is scope for you know sociologists economists to actually think about what what are the implications and there's a there's a great book by jiren ler um who owns the future that is of influence a lot of our thinking on this in the book he talks about the the siren servers you know the big tech companies and three-letter government agencies that M managed to capture all your data and that's basically like a siren

song that we've all Fallen Fallen prey to they have eroded a fundamental human right to privacy and when you look at the foundations of privacy in law you quickly come to um Universal Declaration of Human Rights that enshrines privacy not just as a nice to have not as a privilege but as a fundamental human right attempts by the European Central Bank and other players to turn privacy from a right to a privilege should therefore be viewed with with you know great concern it is a fundamental human right no one shall be subjected to arbitrary interference with his privacy family home or correspondence nor to attacks

upon his honor and reputation everyone has the right to protection of the law against such interference or attacks the problem is that this already makes reference to correspondence meaning communication and that's not an accident because the notion of privacy that that this sort of enshrines goes back to Weston um Alan Weston a sociologist in the 1960s wrote about this and it has influenced a lot of the thinking to this day how lawyers economists think about privacy so he said privacy is the claim of individuals group or institutions to determine for themselves when how and to what extent information about them is communicated to others so he reduces information to

communication that is the first head fake that we are accepting because information has different dimensions there is not just the communication of information there is the storage of information and there's the processing of information there is a reason why there's a difference between information systems and information science and computer science one goes back to Turing the other one goes back to Shannon it is fundamentally philosophically different but our definition of privacy only recognizes the act of communication it's an embodiment of these are not the droids you're looking for now by

accepting it we are accepting this binary choice between disclosing everything or disclosing nothing that is sort of the western idea that if we constrain what is communicated about us we can constrain everything that anybody knows about us and it goes back to Warren and brandise legal Scholars at the end of the 19th century in the US who have written about privacy as the right to be left alone every you know my home is my castle was the idea but it reduces our the control over information to a binary choice and that I think is wrong so the second part of the argument is first part was technology is not

neutral the second part of the argument is data is relational so what we mean with that is the following whenever you aggregate invariably you exert an externality on other people who either have or enough have not contributed to that pool of data whenever you take out a loan repay it or not repay it you create information about other people like you whether they've applied for a loan or not there's nothing they can do about this there's no way you can protect yourself against this kind of inference that comes from the aggregation of information it is something that is

fundamental to the nature of data that is inherent in the in the nature of data salom Von is a legal scholar at Michigan she has a paper specific about this relational nature of data in her argu in in her example there is um two people who have both tattoos one of them um is a gang member uh one of them registers the tattoo at a website for tattoo enthusias web website is being bought by a company that then works with the FBI and other law enforcement so that they create a database of tattoos and gang affiliations the other guy who also has a tattoo who may or may not be a gang member suddenly gets profiled by the

police because of that tattoo why because somebody else created information that this tattoo is associated with a gang affiliation there's nothing you can do against these kind of inferences it is fundamentally embedded in the nature of data and by treating data as something that is absolute that is individual what we would call it despotic for a reason you've seen a second we neglect the relational nature of data darn shimoku who got the Nobel prize in economics this year he wrote a paper a few years ago about these externalities called too much so data because it is relational cannot

be treated as if it was an absolute despotic entity so therefore our hypothesis is wester needs correction you can't talk about privacy in sense of restrictions to communication because it neglects the relational nature of data and therefore it misses what is fundamental to data we should not concern ourselves with privacy that's not a thing not a useful thing anyways we should concern ourselves with control over relational data what kind of data is created how it is processed how it is used and for what purposes by um how often and under what

conditions then is what we should concern ourselves with that would be a much richer set of controls over our data that would be a richer set of Rights all right so uh let me just check in terms of the time okay so I'm I'm going to be quick about this part so when we did this analysis you know we looked at what are the key points in Computing history like who is to blame that was my question who's to blame for all of this uh the tldr is it's the copy function specifically the guys who implemented the first Unix system they messed it up and I'm going to give you an argument why that is the case so when you look at the sort of history of

computing you know there were no users you had the almighty Wizards managing large rooms of computers at their whim data was stored there but you know it wasn't separate from programs before John for noyman came up with ways to deal with that it was enclosed via physical closure Whoever has access to that Mainframe to that big computer they controlled all the data users such as we know them came much later came with the invention of Unix at Babs and there's something to be said about the system being developed at a commercial entity rather than an

academic institution like some of the later oper or multics was for example so when you think about why can't the relational nature of data not be respected it boils down to that whenever you copy data you lose track of where it comes from you don't recall on the operating system any previous operations on a piece of information that is stored on a hard drive you can copy it and you forget all the all the history everybody that has contributed to a database all the individual you know users that have collected their

geolocation information or whatever it is you lose track of all of this because the copy function was implemented as a copy from to but without memory that was a conscious choice of King and Richie and you know all these guys who built the first Unix implementations was not necessary it was not without alternative they they have thought about this other people have implemented different copy functions the elephant OS for example keeps track of every single operation that you do on a file so it's not without alternative I'm going to skip skip this part just to talk a little more

about the source code now we had a similar argument in the late '90s early 2000s around two controls to code and the argument back then was you know make the code open source that protects users control of the data that argument was correct as long as data was stored locally I would argue you know Bill Gates was actually right the Golden Age of privacy and control over data was when we had Microsoft Windows and when we had personal computers it was all downhill from there because at that point in time data was stored largely decentralized not every everything of

course you had large corporations Airlines Banks whatever they always had their big databases of course but personal information what you would you know work with on a day-to-day basis was not in the cloud was not enclosed by Google or Microsoft or Amazon or whoever it was enclosed by you because you had the one and only local copy only with the emergence of the cloud that distinction started to matter so the idea that we should care about who controls the code is only relevant if data is stored in a decentralized way the moment everything gets centralized data gets you know pulled together it doesn't matter as much who controls the code or if you know what the code is supposed to do you can never

con you can never verify that this code is being executed on your data so the open source movement I think they really missed the trick here they missed the analysis of where they are in Computing history and what the existing prevailing information infrastructure means for the battle over control over information and control over over your data all right so you can ask so who won this battle over source code because on the one hand vast majority of devices runs an open source at the same time Microsoft is still doing quite

fine so I would argue that it's because they missed this this trick they they actually lost the battle you can have both you can have World running largely an open source and still having these large siren service data collectors all right because the hobbyists did not anticipate that data can be enclosed even if the code is free they always thought data is going to be distributed and it's locally enclosed and that's not true so I blame them for some of our predicament all right so then the third thing the provence of data if you want to if you want to have control over data you need to know who contributed to it if data is

truly relational if it is not individual if it's not something that you by your very existence and only you create then you need to keep track of who contributed to it and what that is difficult in a word where the copy function has no memory and where the next best thing we created which is the web also has no memory we have cookies and all these these tracking Technologies because there is no history of where you were before on the web it was a conscious design Choice by implementing hypertext as oneway linking know you click on a link and then you go to the next website but the protocol is built such that it doesn't

remember where you come from without additional things like cookies you are never able to overcome these these constraints you're never able to create provence of information without Provence without keeping track of who created the information you can never create property rights it is a fundamental condition for property rights so Tim berners Lee is to blame as well um he's I continue to blame him because he's now doing The self- Sovereign data storage which which is the next bad idea because it again doesn't reflect the relational nature of data all right so the the provence of

data is important we had an alternative there's a wonderful documentary out there by wner herzo called lo and behold reveries of the connected world if you haven't seen it it's on YouTube check it out it's really worthwhile it is about this this battle in the 60s and 70s about hypertext which seems super archaic or and super arcane who would why would we care how hypertex is implemented well we care how it is implemented because it shapes how information is processed how it is stored how we deal with it what kind of Rights we can create to it and these kind of small decisions because technology is not neutral they do matter so he proposed sanadu which was never

finished like wired called it once the greatest vaporware in Computing history uh I don't know if there's still a working version of it but the ideas are valid all right so Ted Nelson was right that's the third hypothesis data needs Providence hypertext should be B directional and the copy function should keep a history of where how a file was processed all right oh we do have examples with information Provence today open source software is one of them in an open source Library you know everybody that road code that you are reusing so you keep track of where individual pieces of code come from um so this Paradigm of reusing existing code that creates Provence in academic

research the convention to site everybody that influenced you're thinking that creates provence of information with digital art and mashups and nfts you do have provence of where individual pieces of information come from so it's not that we haven't seen this right it's just that when it comes to the storage of data we haven't we haven't implemented it all right and that gets me to the next point the despotic data Dominion so now that we've s of accepted you know data is relational technology is not neutral the way we store data doesn't necessarily lend itself to property rights let me drive this point a little bit and talk a little bit about how law and economics struggle to conceptualize data for an

economist data uh ownership boils down to residual control rights that's the core of corporate finance the board exists to decide what should happen with a the company in all the cases that aren't codified in a contract with the manager and contracts are always incomplete that's what we know from har and Moore and there's a Nobel priz just for that inside so property in this in this sense ownership is the sole and despotic Dominion which one man claims and exercise over the external things of the world in total exclusion of the right of any other individual in the universe this is how we conceptualize data as something that is subject to total

uh to Absolute dominion and therefore I would call our existing way of storing data despotic data storage because it has this idea that you have full control over it and steeply rooted in in our economic thinking there's whole theories just built on this idea of incomplete contracts in practice however lawyers tend to think of ownership more in the sense of a bundle of Rights some of which can be reassigned to others there's three dimensions if you want to create property rights so law has long history in that the three dimensions are you need to be able to enclose it you need to capture it you need to be able to declare that this is

yours put up a sign and you need to be able to shoot anybody who you know who tries to trespass to defend your enclosure that is exactly what big tech companies have done they've enclosed your data put up the Declaration that they own it and they sue you if you try to break into it or hack it that's that's how they created property rights and ownership over your data now the problem with this is for economists because in the absolute world you can always copy data if you have access to it therefore it's nonrival that means you can't build normal marketplaces for it it's combinatorial bringing together different pieces of data makes it more less valuable than the individual pieces it's hard for economists to deal with

that so the next hypothesis is that new technology will create new law and new economics because technology is not neutral the way we build it and we design it matters for how we can create laws and how we can create you know legal Frameworks around around the data so how does this road from surom look like now the one thing that I think blockchain actually does and where it's useful for is that it makes data unique you can't create a copy without the system noticing not an authentic copy anyways always you can always create a copy that does not have Provence create

a screenshot or just write down the number on the Block Explorer whatever you want that's not the same as creating a copy with Provence with blockchain that's no longer possible and it's not possible in a decentralized system it's the first time in Computing history that that is the case I think that that's what really makes it unique also how how how can you use that this is where the digital property rights the the digital rights come in the system that we have built uses three things to create data control it uses Machinery to execute code over encrypted data so that everybody can have encrypted data

combine it in a way that doesn't reveal the data to a third party except within the confines of what trusted execution environments can do which has limitations so you take these secure enclaves as a Computing mechanism you record the rights that anybody should have to your data to your pool of data on a system that isn't centrally controlled where you can't be infringed in your rights because nobody can create a copy of them that's a blockchain and you Stitch the two together through an oracle layer because enclaves don't have States blockchains aren't normally used this databases so you need to stitch them together an oracle layer

together with the Byzantine for tolerance gives you the security that The Enclave knows what the state of the blockchain is taken together you get control so in a digital rights token so theide the user story is the following let's say all of us wanted to monetize our geolocation data we could come together form a data pool under the rules that um the median or you know some averages some stati statistics of our pool of our data can be computed but no individual geolocation will be revealed to anybody that's a very specific digital right we could encode this by linking it to an open-source version of the code that executes that implements this right create a smart contract on your

favorite chain that issues tokens that data analysts can buy that reference this piece of code and these tokens whenever you want to redeem them they are sent to the trusted execution environment so uses to oracles to check whether everything is valid then they do the code execution and they can attest to the outside world that this is the code that was run on your data this is what gives you the control so in this system you have a very very different set of Rights for your for your data it's not just read WR execute it's read you know with a blue lens or a yellow yens lens read by a specific person at a specific point in

time a specific number of times so in this system you have this this Data Vault that creates these pools you manage your digital rights we implemented this first on algorand we are moving it to Solana at the moment for variety of reasons um so you create your your digital rights tokens you can trade them or reassign them them on some form of marketplace if you want and if a data analyst here a third party wants to use them they can you know acquire one of your rights execute the code send this right to the to the execution Enclave here it pulls the data does to code execution sends you the

result and in no way in this system can your right be infringed that makes it fundamentally different from the existing infrastructure that we have today and there's a whole range of um sort of technologies that you could use instead of trust execution EnV if you don't like them for whatever reason um that we are currently sort of experimenting with all right so blockchain was this missing piece to move from despotic data control to relational data control and that's I think what makes the the point in time that we're at quite unique because before blockchain we didn't have these technological possibilities so for the first time in

history history we have the tools to fight against this digital despotism so let's go pick a fight thank you for questions like one or two or do we have yes thanks for the speech romes here we talked before uh great um thought provoking presentation so your view of blockchain um you know is basically you're looking at as an immutable Ledger right just for storing data uh you didn't look at the economic perspective of it like crypto economics and in the incentive model and

that was not covered right in your perspective yeah yeah that's true I just assume there is a decentralized ledger whether that's true is another question y thank you I um quick question to the um criticism of The self- Sovereign um data storage um do you think it's just like actually a bad idea or just an insufficient idea um are there in your design no like self-serving data vaults um and it's all relational so it's it's basically we don't need to manage it individually anyways well let me put it this way uh privacies like virginity you can only lose it once and when you lose it

it's very likely you get and that's exactly what self Sovereign data storage does it makes it easy to lose your virginity maybe that's a good thing but I don't like it it could be combined with the relational pool right so it could be you can use it as a first point in a bigger infrastructure but in itself it is insufficient another one last one uh great talk thanks um seems like most us cases around finance and how do you manage wealth and optimize the financial institutions and markets how would you monetize time is I believe it's the most important resource and the scarers for any human on this

planet oh yeah so actually the examples are just because I worked in finance so I know what like for compliance it matters Banks spend lots of money on compliance and gdpr and these things uh Healthcare applications are much more important we did One Pilot with um medical researchers for to help them to combine different medical data for example if you want to find out anti-biotic resistant strains in hospitals they need to combine their data from different hospitals but they are owned by different entities so that's an application for technology like this eventually I I hope that this is a different kind of infrastructure it becomes an alternative to the cloud model that we have and then everything like all our

personal data should be stored in a relational way not in a despotic way cool thank you great thanks