In addition to the samuru.com search engine, they've also created another display of their "Liquid Helium" tech at:
There are some youtube videos about it as well:
Imagine you want to know everything that was ever said about your product, your competitors' products, and comparisons between them.
If you are Apple that is a LOT of stuff. So what if you wanted a machine to do it?
You'd need a copy of the entire Internet. (like a search engine does)
You'd need something that could parse sentences to determine all the features talked about and if they were talked about in a good way or a bad way.
You'd need something to summarize the articles so that humans wouldn't have to flip through every 3000 word review but could still get the highlights. (like our TLDRStuff.com does)
You'd need something to check out the authors to see what their audience reach is, audience demographic, how often they talk about the topic, and how authoritative they are on the subject.
We have all those parts.
We think to build the next generation of search and research engines you have to have all those parts. Link Authority is too easily gamed but putting these parts together we don't need link authority.
Our challenge is making this fast enough to use all the factors, and figuring out how to not penalize brands for searches of their own names since so many brand home pages don't have good information, but shouldn't be number 36 in search results either.
These ideas have been around for a while, and some of them are starting to become more prevalent (phone number recognizers in phones, location recognizers to link to a map, etc.).
Here's an example paper that talks about local context analysis:
http://www.cs.uml.edu/~haim/teaching/iws/tirsaa/sources/ACM_...