4e will be somewhat rules-lighter and more streamlined. 5e will be more rules-heavy. The pendulum will swing back and forth.
The core game experience will remain mostly the same for most groups, but gaming online will become more and more common. (By 'gaming online', I mean playing D&D across an online link of some sort, rather than playing an MMORPG.) Eventually, we'll see a true VR tabletop, which will be quite nice, but that's a long, long way off.
Collectable elements will increase in number and acceptance. The next edition will have a collectable 'grimoire' of spells, with annual volumes collecting all the spells for the year's set in a single volume (in addition to the 'core set' in the PHB, and any spells they publish in the expansions), and collectable magic items under the same model. 5e will extend the idea of to collectable monsters (collected in the annual MM volume - and again in addition to the 'core set' in the first MM). The rules themselves will not be made collectable, and neither will 'big' character building options such as classes (base and PrC), or even feats. The key here is that the collectable elements will add things to the game, but they absolutely will not be required to play... Wizards know that requiring collectable elements will likely kill the game.
Oh, and there won't be a d20 license for the new edition. Instead, Wizards will probably look into favourable but closed licenses with Paizo, Green Ronin, and possibly a few others, to produce key support products that they themselves aren't interested in. There won't be a new SRD, but there won't need to be, as the existing OGL remains valid of course (and d20 now seems to be essentially dead).
(Oh, yes, I almost forgot: as time passes, it will become increasingly apparent that I'm guessing rather wildly. At least 55% of the predictions made here will not come to pass.)