No one ever considers the volume of photos and documents we are now creating. My parents probably took dozens of photographs when I was younger, and kept maybe one or two albums. Nowadays, most people have hundreds if not thousands of photographs and videos. So it seems to me that the converse will actually be true. If you've got a 100 fold increase in the number of documents and photos being created, and only 1% of those make it, you're still preserving the same number of documents. And frankly, I think that's a conservative number.
As some of the other posters here mentioned, with the advent of cloud based services and easy backup systems, retrieval is getting easier. So if storage and retrieval is solved, that leaves format evolution problems. But somehow I doubt in 20 years or even more that JPEG is somehow going to be harder to read than it is now.
A couple of years ago, Kodak's online photo storage service (in the UK at least) shut down. We just barely managed to get copies of the hundreds of pictures we had stored there in time.
Other recent (non-photo specific) examples of vast amount of data disappearing:
Geocities. Yes, large chunks of it was archived last minute. MegaUpload.
And we have Rapidshare on it's way to disappearing.
Cloud services can and will shut down, and it is not at all a given that we manage to preserve the data.
This creates a relentless churn where some proportion of older data disappears every day. All we really can do is to fight to keep the churn rate low enough, because we have no realistic prospect of saving everything all the time.
I don't think this invalidates the point though. In those alone we may have already lost more pictures than were ever taken before 1990 but at the same time the amount we have left is staggering. If you go back 100 years, there are famous people that we have maybe 1 picture of. No matter how sloppy we are with the majority of internet content today, I can't imagine that 100 years from now, they'll only be able to find a couple pictures of Obama and Putin.
But that's not really any different than any other medium. If there exists two copies of a book in different libraries and then one of the libraries burns down it's not any different. I mean we still talk about the library of Alexandria 2000 years later.
As some of the other posters here mentioned, with the advent of cloud based services and easy backup systems, retrieval is getting easier. So if storage and retrieval is solved, that leaves format evolution problems. But somehow I doubt in 20 years or even more that JPEG is somehow going to be harder to read than it is now.