> Tiny self-replicating machines which produce other things to atomic precision
Don't expect it to happen soon - we can't even make big self-replicating machines yet.
We don't even know what machines would look like at such a small scale. Structures behave very differently than they do at human scale; fundamental components like ropes, gears and pulleys might not be possible.
And then there's the problem of programming these machines; you can't exactly put an AVR inside them.
I don't think there's any particular reason to believe that building a large self-replicating machine wouldn't be feasible with modern technology. It's just not useful or cheap, so nobody's done it. NASA studied the idea and I don't think they came up with anything too impossible about it:
From-scratch nanoscale replicators are a whole other beast. However, I think we can say that we know what they would look like, as we have a ton of examples from the natural world. Building ones that do our bidding is a bit tougher, obviously.
A machine that could build faithful copies of itself using only materials from the natural environment (i.e. not this http://en.wikipedia.org/wiki/File:MolecubesInMotion.jpg) would be absolutely revolutionary and would essentially be a completely alien and artificial life (mutation not being especially difficult to include, and probably difficult to prevent). It would change the world, because once you have replication you can piggy back just about any other ability on top of it and do it on a massive scale, no matter how inefficient at the individual level.
I will gladly pay you $100 billion for a self-replicating machine.
Don't expect it to happen soon - we can't even make big self-replicating machines yet.
We don't even know what machines would look like at such a small scale. Structures behave very differently than they do at human scale; fundamental components like ropes, gears and pulleys might not be possible.
And then there's the problem of programming these machines; you can't exactly put an AVR inside them.