Do tech companies really store unbounded amounts of information, like keyboard strokes, from their users? For me this seems hard to believe, at least over the medium to long term.
Storing data is not free and if you plan to use this data for anything it can definitely be not cheap (i.e. cold vs hot storage). When you extrapolate this to the scale of a service like Facebook or Google it becomes insanely expensive. So there would have to be some pretty clear monetary payoff to justify this expenditure. I understand the worries of privacy advocates, but I don't understand the economic incentive of keeping all this data. The only way it could possibly make sense is if it were highly aggregated, but at that point you nullify many of the privacy concerns.
> Storing data is not free and if you plan to use this data for anything it can definitely be not cheap
I think there are a lot of companies that are not running out of money and are holding on to a lot of data that might, someday, be valuable. Look at Google's features like spelling correction or search term suggestions - they likely used huge troves of semi-anonymous user input to develop and support them.
> When you extrapolate this to the scale of a service like Facebook or Google it becomes insanely expensive
This is just not true. Backblaze is putting their storage cost at ~$0.035 a gigabyte[1], which means you can store a megabyte of data for every human on earth for a bit under $300k a year (about 8 petabytes). Big companies probably can get even lower. This isn't...a negligible cost, but it's very normal at that scale. Google only has ~4.5 billion users and most companies have orders of magnitude fewer.
> Look at Google's features like spelling correction or search term suggestions - they likely used huge troves of semi-anonymous user input to develop and support them.
There's a big difference between data which is directly tied to PII and data which is held in aggregate in terms of privacy. I'm not arguing there can't be leakage here, but it certainly blunts many of the more severe privacy implications. Conflating these two is more sensationalist rather than useful in terms of honing in what is ok vs what is not.
> which means you can store a megabyte of data for every human on earth for a bit under $300k a year
Sure, I mean as a slippery slope you could also write this data to paper and keep it indefinitely at a very cheap price. My point here is more: if you plan on using the data, it becomes more and more expensive as your access patterns change. This also has privacy implications because I would imagine the easier the data is to access in raw form the higher the potential privacy cost to the user. If all they are doing with these key strokes is recording somewhere that you might be interested in Corgis and German Shepards based on your keystrokes, as opposed to something more detailed like an accidental paste of your password, I think that changes the conversation.
> My point here is more: if you plan on using the data, it becomes more and more expensive as your access patterns change.
I don't think this is true either. Google does not need to keep much of its data in hot storage to use it effectively: their ML products can be periodically trained / updated, their search can be iteratively updated with each crawl, etc. Sure, it would be expensive to keep all user data from all sources in hot storage all the time - but it's not needed. The idea that you...would happen upon some new question you hadn't though of before and need to get the answer immediately is just false. Instead, you make regular updates to a model and periodically run your corpus through that model.
You're welcome to request your data from Facebook and see if they have this data (assuming they are following the law, which I assume they are). I've requested my data from various companies and they did not have anything even close to this level of detail. That makes me think that if they do store data such as keystrokes or mouse clicks or whatever, it's being stored in a highly aggregated way, which doesn't bother me very much. But I also imagine one of the reasons they are storing data in an aggregated way is: 1) they don't need it at an individual level and/or 2) it's not cost effective.
Storing data is not free and if you plan to use this data for anything it can definitely be not cheap (i.e. cold vs hot storage). When you extrapolate this to the scale of a service like Facebook or Google it becomes insanely expensive. So there would have to be some pretty clear monetary payoff to justify this expenditure. I understand the worries of privacy advocates, but I don't understand the economic incentive of keeping all this data. The only way it could possibly make sense is if it were highly aggregated, but at that point you nullify many of the privacy concerns.