A few reasons:

- Python3 makes some concepts, like unicode, division, the print function, etc, simpler to understand.

- Almost all major packages (only scrapy doesn't that I can think of), including the scientific stack, are compatible with 3.

- The trend towards 3 seems to be accelerating. 2 isn't going away anytime soon, but I'd like this content to be relevant for some time.

Ultimately, they aren't that different, and I may have a section listing out what you need to do to switch between them.

Because Python3 is the current version of python? If you're starting new projects, there are very few valid reasons not to use it.

A bunch of data sciene material, in particular, is written in Python 2.x, and it wouldn't be bad to teach Python2 either. You could always use the _future_ command for newer modules as well