Immediately after a difficulty change, it is nearly impossible to predict what the difficulty will be after the next change (2016 blocks later). Miners could begin shutting down their rigs, new ASIC could come online, someone could discover a weakness in SHA-256, there's just no way to know. Any number anyone gives you is simply a random guess.
As more blocks are solved, we can begin to predict what the next difficulty will be with increasing accuracy. The more blocks we've seen since the previous adjustment, the more information we have about how fast blocks are being created and what the average time is between blocks. By the time we've seen 2015 blocks since the last difficulty change, we can be nearly certain of what the new difficulty is about to be, but still have no idea what the next difficulty will be 2016 blocks after that.
Therefore, while we can say that the difficulty will change approximately every 2 weeks, we really can't predict with accuracy how fast it will increase over the next few months, or years.
If you can predict how much hashing power will be taken offline (CPU, and GPU miners giving up), and how much new hashing power will come online (new ASIC being added to the system), then you could perhaps calculate an estimated difficulty based on those predictions, but I haven't seen any reliable predictions of either of those pieces of information anywhere.
It isn't really useful, but you can set an upper limit on difficulty.
https://en.bitcoin.it/wiki/TargetA single retarget never changes the target by more than a factor of 4 either way to prevent large changes in difficulty.
