This topic comes up every once in a while and I've done some rudimentary research on world record progressions that has resulted in me saying it will probably not happen before the late 2020s and would very likely take longer. Now, someone has apparently done a little more comprehensive look into it. Here are the results.
Probably what most people are interested in:
Using times from 1960, the open squares suggest it will take 12-13 years to break 2 hours assuming a ~20 sec reduction per year. If times from 1980 are used the filled squares suggest it will take 25 years assuming a ~10 sec reduction per year.
Of course, this assumes a linear progression. As one can see by the quote above, the progression is not linear. If it were, we'd see a consistently yearly reduction and the times from 1960 forward would produce a number at least significantly similar to times from 1980 forward. In other words, in the future, even a reduction rate of ~10 seconds per year will likely not be maintained.
In short, it would seem like a very unlikely scenario that we will see someone break 2 hours in 12-13 years. More likely would be 25 years but it could be even longer.
I have a rudimentary spreadsheet I came up with after Geb broke the 2:04 barrier that shows very similar numbers. Since the 2:10 barrier was broken in 1967, the average improvement has been ~1:00 every 6 years or ~10 seconds per year as the paper linked mentioned as the average improvement since 1980. Based on this progression, I calculated that the 2:00 barrier would be broken in 2032 or 24 years after Geb broke the 2:04 barrier.
Of course, humans are not machines. We break from statistical norms all the time. However, we typically don't stray too far from statistical norms.
If I have a chance, I'll clean up this spreadsheet and post some form of it online. Maybe it would make for an interesting article.