Although I'm not sure how much sense it makes and how useful it might be, I find space colonization cool. I think, rather than building space ships and sending humans to space to build space colonies it might be better to send self replicating machines and build colonies and space ships and then send humans there and use them.
So the question is, how far are we from building self replicating machines? 
There's a lot of automation already, autonomous car driving, lights out manufacturing  and even one of the most complex tasks, production of integrated circuits is striving for lights out.
The Mars rover Curiosity can be told to move to a certain place and then does so autonomously.
OTOH a recent development was a robot that is able to find a socket and plug itself in to charge it's batteries ... with a 93% success rate. It appears a bit clumsy, as it doesn't use vision to find sockets, as they come in different shapes and colors (and the lights may be out), so it senses the electric field.  This makes me wonder how much we still have to work on real world visual processing and understanding.
The replicators don't have to be completely autonomous. They can still communicate with human and report difficulties and be updated, given help or new tasks (build habitats, space ships, satellites, mine asteroids and transport stuff, recycle space junk, etc).
How long could it be until we can assemble what we already have and create the missing links to get such machines? 10 years?
Bernhard