Google's Knowledge Graph is pretty good at telling you who was the 37th president of the US, or what the square root of 342345 is. Ask it more complex questions, like "why does the sun set at night?" and it'll still send you off to find the answer yourself. Next week in New York, Google researchers will present a paper on its "Knowledge Vault," which Kevin Murphy of Google Research, describes as "the largest repository of automatically extracted structured knowledge on the planet." Knowledge Vault applies machine learning (unlike Knowledge Graph which is an extension of community supported tools) to automatically trawl webpages, assimilating their facts, information and connections therein. Not only does this mean it's faster, it can continually grow and update itself. The net result will be a huge database of knowledge, the likes of which would have been unimaginable just years ago.
The real-world implications are that services like Google Now (or Siri et al) could get a huge boost in smarts -- tapping into a much deeper well of understanding, knowing what, how and why things are related. This could lead to much more intelligent web services, or truly explode any limits of augmented reality ("ah, you're in Berlin, and sent an email last week about museums, perhaps you want to visit the Museum Berggruen"). Unsurprisingly this comes at a privacy cost; analysts are expecting Google to leverage services like Gmail that contain your data (plus the data that's public/online) bundled in with the rest of the world wide web. While there are no timelines on when we might see this implemented in live services, don't be surprised when it's not just your proximity to Sir Bacon that freaks you out, but how and why you're so close, plus favorite films you have in common, and where you can go and see them locally.
Update: Google has contacted us with with some clarification on how the project will work. Most importantly, the analysts are wrong, and this system wouldn't use any of your personal data. Secondly, "Knowledge Vault" is one of many efforts Google is working on to better understand text and what we're searching for, and has been ongoing for some time. We've also included a link to the paper in "More coverage."