Microsoft appears to be in the business of trying to build a fully-functioning virtual assistant. If they're solving that problem with big data, they need a full stream of the user's behavior to operate on. But the tradeoff---an actual, working virtual assistant---could very well be worth it to people for whom that personal information is not worth hiding from a corporation with no vested interest in undermining their customer base by irresponsibly divulging that data.
If anything's sad, it's that we live in a world where some people do not have the freedom to treat their private info so loosely. That's a true tragedy, because there will increasingly be technologies they can't take advantage of.
edit: just wanted to mention that when I first installed ubuntu I forgot to disable the feature, performed some very innocuous searches, and was shown loads of porn results from Amazon. That alone makes the feature completely worthless and frankly dangerous (depending on the environment that you're using ubuntu in).
1.) No Internet access (the default)
2.) Route through Tor
3.) Route via VPN to a server at home and out through my broadband connection (I trust my broadband ISP slightly more than my mobile network provider). This protects me on untrusted WiFi networks too.
[edit] I wish I didn't have to do all this, but the smartphone OS market doesn't give me non-leaky OS options in the same way that the Desktop market does.
Luckily Cyanogenmod supports my phone. I will attempt the installation soon. Any piece of advice, or something you would like have known when at my situation?
edit: nm, based on the other downvotes in this thread I'm guessing zealots at this point.
I assume the same sort of stuff is happening on Ubuntu today, but I stopped using Ubuntu on the desktop/laptop when this came out and started using Debian instead.
but regardless, they've lost too much respect/credibility for including it in the first place, so I wouldn't trust them in general.