Google and Bing are already quite good at predicting what you're thinking and giving you relevant suggestions. But it's slow and the predictive algorithms just aren't quite good for actual real-time suggestions. So a graduate student from Cornell wants to speed things up considerably, making things potentially awesome, yet incredibly creepy.
We all like our personalized suggestions when shopping or even when searching for things. It legitimately helps us, despite the privacy concerns related to it. But those results don't show up instantaneously. There's a significant amount of background work going on that links your actions to those personalized results. Wenlei Xie has come up with an algorithm that could potentially speed things up to near real-time.
Search engines and their underlying suggestions generally use a weighted node graph which is examined analyzed to see just how appropriate the suggestion is based on years of collecting and correlating information. The problem that Xie has found is that there's just so much information to walk through, that it's incredibly slow. So to make it faster and more relevant even sooner, he's proposed simplifying those graphs. In essence they're assembling only the most pertinent information, and discarding a lot of the fluff, to make it quicker.
They've already tested out their unique method on a set of scholarly databases and even a blog search system which resulted in it working up to five times faster than the normal methods. In time this could help to reduce the amount of time you take searching for important things, and might reduce the number of letters you type before the engine knows what you want.