GPS doesn't "work" in the same way that he is referring to. For the purpose of identifying a point on the globe, with line-of-sight to the sky - GPS works fine.

But, what is really being talked about here is much deeper spatial awareness based on a range of sensor and relational data modeling.

They are using GPS, direction, lighting, sound and finally - actual images to determine all the spatial attributes of where you are, what your surroundings are and whom else is in your vicinity -- even if you don't know them.

This is stunning.

They will be able to know which parties are groups - where they are in relation to one another etc.

If they are already using all those data-points. I am absolutely positive they are will have audio voice recognition - keyword recognition in the future as well.

There are incredible uses for this, such as "here were some highlights of the parties conversations"

What we NEED is for someone to sniff the backhaul from the client and lets look at exactly what data is leaving our devices.