Exploring algorithmic reputation and governance. replabs.xyz
Exploring algorithmic reputation and governance. replabs.xyz
Subscribe to Oliver Klingefjord
Subscribe to Oliver Klingefjord
<100 subscribers
<100 subscribers
Share Dialog
Share Dialog
I recently read "The Sovereign Individual", a book cherished by high-ups in the crypto scene like Balaji Srinivasan and Peter Thiel. Thiel wrote the preface to the book, in which he summarizes the current state of tech as an ideological struggle between AI and crypto. On one side, you have CCP and centralized control. On the other, the crypto world inhabited by sovereign individuals.
In camp crypto you find web 3 projects ranging from Blockchains to Tim Berner Lee's Solid. "You should own your data" is the mantra, and “you shouldn't be tracked by anyone without your consent”. On the other side is China in the east, and Google and Facebook in the west – gigantic, centrally-owned data collectors who train AI models to predict and modify its users' behavior.
The sovereign individuals lament the state we find ourselves in – the internet, originally a radically distributed system, has converged into a few behemoths. In the face of this, it is easy for anyone with a love of freedom to side oneself on team crypto. Why wouldn't you want to be in control of your data? Why wouldn’t you want to be a sovereign individual?
Let's take a step back;
What exactly is your data? Your ID number for example – is that yours? If no government agency knew that your ID number was indeed connected to you, what would you do with it? It would be an arbitrary string of numbers to anyone but you – useless. Its value comes from the relations that collectively attest to this string of numbers being your ID number.
What about your genetic code? Well, it contains information about your parents – disclosing it without their consent could be seen as a breach of their right to control their data. Likewise, your email inbox contains information from those who sent their messages to you. Your taste in music is influenced by scenes around you. In short, your digital identity is spread across a network rather than being a discrete set of information.
I recently read "The Sovereign Individual", a book cherished by high-ups in the crypto scene like Balaji Srinivasan and Peter Thiel. Thiel wrote the preface to the book, in which he summarizes the current state of tech as an ideological struggle between AI and crypto. On one side, you have CCP and centralized control. On the other, the crypto world inhabited by sovereign individuals.
In camp crypto you find web 3 projects ranging from Blockchains to Tim Berner Lee's Solid. "You should own your data" is the mantra, and “you shouldn't be tracked by anyone without your consent”. On the other side is China in the east, and Google and Facebook in the west – gigantic, centrally-owned data collectors who train AI models to predict and modify its users' behavior.
The sovereign individuals lament the state we find ourselves in – the internet, originally a radically distributed system, has converged into a few behemoths. In the face of this, it is easy for anyone with a love of freedom to side oneself on team crypto. Why wouldn't you want to be in control of your data? Why wouldn’t you want to be a sovereign individual?
Let's take a step back;
What exactly is your data? Your ID number for example – is that yours? If no government agency knew that your ID number was indeed connected to you, what would you do with it? It would be an arbitrary string of numbers to anyone but you – useless. Its value comes from the relations that collectively attest to this string of numbers being your ID number.
What about your genetic code? Well, it contains information about your parents – disclosing it without their consent could be seen as a breach of their right to control their data. Likewise, your email inbox contains information from those who sent their messages to you. Your taste in music is influenced by scenes around you. In short, your digital identity is spread across a network rather than being a discrete set of information.
“We should think of ourselves as the intersection of the social groups we inhibit, and what defines our individuality is that no other individual is part of precisely the same collection of social groups” – George Simmel
It's hard to think of anything that's truly your data. Also, what use would your data be without context? Data is not divisible. A single datapoint is worthless. The value of a dataset, as in social networks, is proportional to the number of relations – it grows exponentially with size.
Unfortunately, this means that data favors centralization. Five unrelated datasets of a million social media users each is orders of magnitude less interesting than one dataset of five million densely connected users.
This questions the thesis of the sovereign individual projects and you owning your data profile – what use would this profile be if there was no other profiles to compare it with? What use would it be to store information about us in isolated buckets if the buckets didn't add up to something meaningful?
Data is power in the Information Age, and as discussed above, the power grows exponentially in proportion to the size. Thus, sovereign individuals with their isolated data buckets would be practically powerless in the face of the rising force in the east. Will we, in a world of Urbit and Solid, be able to respond to the crises of tomorrow?
Covid was an interesting litmus test for government efficacy in responding to crisis. Many countries in the west, like the US, failed the test horrendously: -3.5% GDP growth and 224 deaths for every 100.000 citizen in 2020. The UK fared worse, at a -9.9% GDP growth and 208 deaths per 100.000 citizens. Even Switzerland, universally cherished as a stable, competent country, suffered -2.5% GDP growth and 130 deaths per 100.000 citizens. On the other hand, Taiwan experienced a positive GDP growth of 3.1% with only 4 deaths for every 100.000 citizens.
This was partly made possible due to the Taiwan’s advanced ICT infrastructure. Taiwan was able to contract trace and perform rapid data-driven decision-making. Perhaps this success was what inspired the west to clumsily tiptoe into surveillance and vaccination QR codes. Regardless – data is power, and the power granted by data scales exponentially with its size. States and organizations that are able to operate using big data are able to outcompete those who don't.
Regardless of what you think of the nation states in the west – there is a certain operational capacity that comes with centralized data that should not be ignored. At the same time, it is true that centralized power is a dangerous prospect – a dictator might be useful in a crisis, but installing one tend to be a one-way street. Freedom is hard to earn and easy to loose.
There is a subfield within crypto that tries to address this problem by allowing people to get paid for their data. Big organizations can choose to sell access to their data, and in that way, we could create a world where we can learn from each other across silos.
This is one step in the right direction, but ultimately it doesn't change the aforementioned power dynamics too much – data would still be a force of centralization. Instead, big companies get the option to convert some of its data power into monetary power. If this power equation did not work out for them, and the companies that buy their data threaten their hegemony, there wouldn't be any incentive for them to do this in the first place.
Of course, if the data is sold to a company from a different industry, no power is threatened and everyone is happy. In this case, however, Ocean Markets is yet another way for megacorps to further capitalize on their data, leading to yet more centralization of power.
Is there another way out of this conundrum? Do we have to opt for either centralized power or decentralized powerlessness?
I know this sounds like a privacy nightmare – I can hear screams of cypherpunks in the back of my head. But stay with me for a second:
Data grows in power exponentially with its size. Data as a public good means that data monopolies wouldn't be able to form and cause lock-in effects, making it infeasible for new startups to enter the space. Big tech, startups and NGOs would all share the same playing field. However, the free world would still be able to exercise the power that comes from big data. If data is never allowed to aggregate but lives in isolated buckets, the decentralized world will not be able to put up much resistance in face of AI superpowers.
It also would get us out of the tricky question of who owns data. If a datapoint is interpersonal, which person should be in charge of storing that datapoint? In case you want to share an interpersonal datapoint, what happens to others who might have their identity compromised by it?
Of course, data as a public good doesn’t mean that all data ought to be universally accessible.
In the original cypherpunk manifesto, Eric Hughes defines privacy as "the power to selectively reveal oneself to the world". You could view privacy as the power to decide which part of yourselves are not yet ready for public display, and which parts of you aren't.
The networked nature of data, however, makes it hard to reveal "a part of you" without simultaneously revealing information about others. As mentioned above, our digital selves are deeply enmeshed with one another.
This means that its hard to think of privacy as a binary switch (in the way blockchains do – any data is either publicly available on-chain or stored privately in a wallet). We ought instead to think of privacy as the ability to "live privately and selectively reveal parts of yourselves to others". This value can be respected by services, organizations and companies to various degrees.
One alternative approach to privacy is to keep a collective record of how entrusted organizations are. An organization entrusted by people to follow good practices would be allowed to access a lot of data, whereas organizations that abuse privacy would quickly be restricted. In the same way that proof-of-stake punishes bad actors by removing their staked funds, you could punish bad actors that abuse your right to privacy by removing their staked reputation.
There are many potential solutions, involving combinations of federated learning, homomorphic encryption and ZK validation. I don't have a concrete vision of how this could play out, but I'm convinced that in order to move away from centralized data ownership, we need to start exploring this space and recognize that we need to be able to learn from each other’s data somehow.
I've talked about data and power abstractly so far – let's make it concrete. My personal aim with regards to public data is to create systems for legitimization that are better than today’s centralized algorithms. In the web 2 world we find ourselves in, big tech control these algorithms, and consequently, the narrative. However, the sovereign individual world doesn't offer a viable alternative.
Google PageRank was a system of legitimacy created by crawling public data (publicly available webpages) and ranking content by its relational density. For all its imperfections, it is something that would not exist in a world of sovereign individualists – its hard to imagine a search engine if every webpage read required consent from the publisher.
The sovereign individual idea sometimes feel like the digital equivalent of retreating into the woods and shutting yourselves into a barn with a shotgun, ready to shoot anything that moves. Merely "running away" from the woes of web 2 are not going to solve them. We need better systems for legitimacy, and better systems for surfacing what's good, meaningful and beautiful. And we could have them, if we dared to think outside the block.
Like this article? Consider buying me a coffee ☕️
“We should think of ourselves as the intersection of the social groups we inhibit, and what defines our individuality is that no other individual is part of precisely the same collection of social groups” – George Simmel
It's hard to think of anything that's truly your data. Also, what use would your data be without context? Data is not divisible. A single datapoint is worthless. The value of a dataset, as in social networks, is proportional to the number of relations – it grows exponentially with size.
Unfortunately, this means that data favors centralization. Five unrelated datasets of a million social media users each is orders of magnitude less interesting than one dataset of five million densely connected users.
This questions the thesis of the sovereign individual projects and you owning your data profile – what use would this profile be if there was no other profiles to compare it with? What use would it be to store information about us in isolated buckets if the buckets didn't add up to something meaningful?
Data is power in the Information Age, and as discussed above, the power grows exponentially in proportion to the size. Thus, sovereign individuals with their isolated data buckets would be practically powerless in the face of the rising force in the east. Will we, in a world of Urbit and Solid, be able to respond to the crises of tomorrow?
Covid was an interesting litmus test for government efficacy in responding to crisis. Many countries in the west, like the US, failed the test horrendously: -3.5% GDP growth and 224 deaths for every 100.000 citizen in 2020. The UK fared worse, at a -9.9% GDP growth and 208 deaths per 100.000 citizens. Even Switzerland, universally cherished as a stable, competent country, suffered -2.5% GDP growth and 130 deaths per 100.000 citizens. On the other hand, Taiwan experienced a positive GDP growth of 3.1% with only 4 deaths for every 100.000 citizens.
This was partly made possible due to the Taiwan’s advanced ICT infrastructure. Taiwan was able to contract trace and perform rapid data-driven decision-making. Perhaps this success was what inspired the west to clumsily tiptoe into surveillance and vaccination QR codes. Regardless – data is power, and the power granted by data scales exponentially with its size. States and organizations that are able to operate using big data are able to outcompete those who don't.
Regardless of what you think of the nation states in the west – there is a certain operational capacity that comes with centralized data that should not be ignored. At the same time, it is true that centralized power is a dangerous prospect – a dictator might be useful in a crisis, but installing one tend to be a one-way street. Freedom is hard to earn and easy to loose.
There is a subfield within crypto that tries to address this problem by allowing people to get paid for their data. Big organizations can choose to sell access to their data, and in that way, we could create a world where we can learn from each other across silos.
This is one step in the right direction, but ultimately it doesn't change the aforementioned power dynamics too much – data would still be a force of centralization. Instead, big companies get the option to convert some of its data power into monetary power. If this power equation did not work out for them, and the companies that buy their data threaten their hegemony, there wouldn't be any incentive for them to do this in the first place.
Of course, if the data is sold to a company from a different industry, no power is threatened and everyone is happy. In this case, however, Ocean Markets is yet another way for megacorps to further capitalize on their data, leading to yet more centralization of power.
Is there another way out of this conundrum? Do we have to opt for either centralized power or decentralized powerlessness?
I know this sounds like a privacy nightmare – I can hear screams of cypherpunks in the back of my head. But stay with me for a second:
Data grows in power exponentially with its size. Data as a public good means that data monopolies wouldn't be able to form and cause lock-in effects, making it infeasible for new startups to enter the space. Big tech, startups and NGOs would all share the same playing field. However, the free world would still be able to exercise the power that comes from big data. If data is never allowed to aggregate but lives in isolated buckets, the decentralized world will not be able to put up much resistance in face of AI superpowers.
It also would get us out of the tricky question of who owns data. If a datapoint is interpersonal, which person should be in charge of storing that datapoint? In case you want to share an interpersonal datapoint, what happens to others who might have their identity compromised by it?
Of course, data as a public good doesn’t mean that all data ought to be universally accessible.
In the original cypherpunk manifesto, Eric Hughes defines privacy as "the power to selectively reveal oneself to the world". You could view privacy as the power to decide which part of yourselves are not yet ready for public display, and which parts of you aren't.
The networked nature of data, however, makes it hard to reveal "a part of you" without simultaneously revealing information about others. As mentioned above, our digital selves are deeply enmeshed with one another.
This means that its hard to think of privacy as a binary switch (in the way blockchains do – any data is either publicly available on-chain or stored privately in a wallet). We ought instead to think of privacy as the ability to "live privately and selectively reveal parts of yourselves to others". This value can be respected by services, organizations and companies to various degrees.
One alternative approach to privacy is to keep a collective record of how entrusted organizations are. An organization entrusted by people to follow good practices would be allowed to access a lot of data, whereas organizations that abuse privacy would quickly be restricted. In the same way that proof-of-stake punishes bad actors by removing their staked funds, you could punish bad actors that abuse your right to privacy by removing their staked reputation.
There are many potential solutions, involving combinations of federated learning, homomorphic encryption and ZK validation. I don't have a concrete vision of how this could play out, but I'm convinced that in order to move away from centralized data ownership, we need to start exploring this space and recognize that we need to be able to learn from each other’s data somehow.
I've talked about data and power abstractly so far – let's make it concrete. My personal aim with regards to public data is to create systems for legitimization that are better than today’s centralized algorithms. In the web 2 world we find ourselves in, big tech control these algorithms, and consequently, the narrative. However, the sovereign individual world doesn't offer a viable alternative.
Google PageRank was a system of legitimacy created by crawling public data (publicly available webpages) and ranking content by its relational density. For all its imperfections, it is something that would not exist in a world of sovereign individualists – its hard to imagine a search engine if every webpage read required consent from the publisher.
The sovereign individual idea sometimes feel like the digital equivalent of retreating into the woods and shutting yourselves into a barn with a shotgun, ready to shoot anything that moves. Merely "running away" from the woes of web 2 are not going to solve them. We need better systems for legitimacy, and better systems for surfacing what's good, meaningful and beautiful. And we could have them, if we dared to think outside the block.
Like this article? Consider buying me a coffee ☕️
No activity yet