Maybe it won't change at all in the end. The situation is not much different than, say, commercial software: You don't expect paid programmers to release their code in the open as long as they can take advantage of it for making money, but when that stops, they do. Similarly, researchers will use it for funding, promotions, prestige etc. However, the problem with academics is that even "archived" science remains closed, not just the papers, but also the raw data. So it would be a big step forward regardless. We need a worldwide repository for scientific exchange. The web is ready with open arms, but researchers are not.