[ExI] File survival and the transparency of the future

[Anders Sandberg]

The backup issue brings up another thing that has been in my mind 
recently: what is the average lifespan distribution of data files that 
you don't care about?

Obviously most files are fairly recent - both because they stem from the 
latest computer installation, because the space is growing and hence 
more files are created now than in the past, but also because files tend 
to disappear due to bit rot, software obsolescence, accidental and 
deliberate deletion and so on. However, really old files survive by 
being backed up - they are close to things you care about or think you 
care about. And as storage space grows they have a decent chance of 
hanging on.

Empirically (i.e. looking at the slashdot discussion about the topic) 
most normal people (on slashdot) have files going back about ten years. 
Some oldtimers boast about their 80's files, but most of them have been 
deliberately preserved. And we all know the sad story about the Apollo 
tapes: software and media obsolescence bite hard.

This issue is interesting because I have been thinking about long-term 
storage of data that acts as surveillance data on the present. All those 
logs, geotags, emails and whatnot that we produce can with the right 
technology be used to infer things about the present. Future tech will 
make this data more inferentially promiscous: we will be able to figure 
out things that are entirely nonobvious today from apparently innocent 
files. I doubt this is useful for the reconstructive upload scenario, 
but it is in any case relevant for analysing just how transparent we are 
to the future.


-- 
Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University