5 ms·

Your second point is the critical one, and I'll add to it — there must be an incentive to decentralize, and it must outweigh the numerous incentives to re-centr

by mLuby 8y ago

Your second point is the critical one, and I'll add to it — there must be an incentive to decentralize, and it must outweigh the numerous incentives to re-centralize. We need to centralize around protocols/standards and decentralize ownership of data. If somehow each piece of data on a user were exponentially more expensive to keep, that might be possible. But that would be a very strange law…

tareqak 8y ago
Yeah, it probably more realistic (and possibly even ideal in its own way) to assume that obtaining the desired equilibrium will always be an adversarial process. Having proven ways to measure the dominance of each actor or having canaries / alerts is important, but not enough: there has to be ways to measures detect if anyone has subverted, gamed, or compromised the measures themselves.
lovemenot 8y ago
>> If somehow each piece of data on a user were exponentially more expensive to keep, that might be possible. But that would be a very strange law… Dave and I agree with your diagnosis, and we would like to propose a solution. Who is Dave? Well you know him, and I know that you know him and I know that when we both talk about Dave we must be talking about the same person. (mLuby I probably don't really know you, this is just a hypothetical). A casual observer could not approach our common understanding by adding more data. There's so many Daves after all. In fact they'd need to subtract all but the right data. So the solution is a protocol in which increased data adds noise faster than it adds signal. Such a protocol requires 1000s times more plausible yet incorrect noise for every signal. Digital chaff.
- mLuby 8y ago
  I don't understand. If Eve (who we don't know) adds data like location, age, interests to her profile on our Dave, people will want to access her profile on Dave. This creates an incentive for Eve to aggregate as much data on as many people as possible. Are you saying that nobody can know which Dave we're talking about unless we identify him as Dave-with-cell-5551234567? I don't think that's true, since human social circles are pretty easy to figure out; see the humorous hypothetical collection of data about US Founding Fathers. http://www.newenglandhistoricalsociety.com/phone-spying-paul-revere-and-colonial-social-networking/ http://www.newenglandhistoricalsociety.com/phone-spying-paul...
  - lovemenot 8y ago
    Yes. That is what I am saying. And I agree that the protocol must be so designed that more data does not increase specificity, but rather increases noise.
pymai 8y ago
the SafeNetwork seems like the best decentralised project in that respect. the more resources you share with the network (cpu, hard drive space) the more money you earn. that should more than enough incentives for most poeple! there are a few other things too like only being charged to upload files once vs paying a subscription for the rest of your life, or only needing one password for all sites.

by mLuby 8y ago

tareqak 8y ago

Yeah, it probably more realistic (and possibly even ideal in its own way) to assume that obtaining the desired equilibrium will always be an adversarial process. Having proven ways to measure the dominance of each actor or having canaries / alerts is important, but not enough: there has to be ways to measures detect if anyone has subverted, gamed, or compromised the measures themselves.

lovemenot 8y ago

>> If somehow each piece of data on a user were exponentially more expensive to keep, that might be possible. But that would be a very strange law… Dave and I agree with your diagnosis, and we would like to propose a solution. Who is Dave? Well you know him, and I know that you know him and I know that when we both talk about Dave we must be talking about the same person. (mLuby I probably don't really know you, this is just a hypothetical). A casual observer could not approach our common understanding by adding more data. There's so many Daves after all. In fact they'd need to subtract all but the right data. So the solution is a protocol in which increased data adds noise faster than it adds signal. Such a protocol requires 1000s times more plausible yet incorrect noise for every signal. Digital chaff.

mLuby 8y ago
I don't understand. If Eve (who we don't know) adds data like location, age, interests to her profile on our Dave, people will want to access her profile on Dave. This creates an incentive for Eve to aggregate as much data on as many people as possible. Are you saying that nobody can know which Dave we're talking about unless we identify him as Dave-with-cell-5551234567? I don't think that's true, since human social circles are pretty easy to figure out; see the humorous hypothetical collection of data about US Founding Fathers. http://www.newenglandhistoricalsociety.com/phone-spying-paul-revere-and-colonial-social-networking/ http://www.newenglandhistoricalsociety.com/phone-spying-paul...
- lovemenot 8y ago
  Yes. That is what I am saying. And I agree that the protocol must be so designed that more data does not increase specificity, but rather increases noise.

pymai 8y ago

the SafeNetwork seems like the best decentralised project in that respect. the more resources you share with the network (cpu, hard drive space) the more money you earn. that should more than enough incentives for most poeple! there are a few other things too like only being charged to upload files once vs paying a subscription for the rest of your life, or only needing one password for all sites.