[ExI] data are lost to science at 'astonishing rate'

[Anders Sandberg]

On 2013-12-19 20:01, spike wrote:
>
> My point is that as science loses data at an appalling rate,
> history loses data even more quickly.  The libraries of Alexandria are
> burning furiously to this day, with far more fuel than in 642 AD.

However, cheap copying also makes things far more resilient. If you have 
N copies of something and probability p of something surviving, you 
should expect Np surviving copies, and only (1-p)^N probability of them 
all getting wiped out. This means that on average, we should expect the 
future to have access to Justin Bieber and 50 Shades of Grey. But also 
copies of Nature (N=53,000).

The real problem is when there are few copies, like for scientific data. 
I scared a representative of one of the big brain projects by asking 
what they will do with their giant database after the end of the 
project. The Millennium N-body simulation has AFAIK already been deleted.

Centralization may also be a problem: for all purposes ScienceDirect is 
one copy, even if it runs on the cloud. Were the company to fold it 
might suck down all its scientific papers. Sure, individual scientists 
likely have their manuscripts and they might exist in email backups, but 
reconstructing them would be tough - especially as the Vines paper 
pointed out that even finding researchers after a few years is tricky.




-- 
Anders Sandberg
Future of Humanity Institute
Oxford University