Researchers at MIT and Carnegie Mellon are developing programmable matter: material consisting of tiny machines that can be reconfigured into many different shapes:
How can a material be intelligent? By being made up of particle-sized machines. At Carnegie Mellon, with support from Intel, the project is called Claytronics. The idea is simple: make basic computers housed in tiny spheres that can connect to each other and rearrange themselves.
…
Wach particle, called a Claytronics atom or Catom, is less than a millimeter in diameter. With billions you could make almost any object you wanted.
The concept sounds like a macroscopic version of nanotechnological utility fog. The image is of the most up to date Catom, which is still in the centimetre size range.
The challenges and opportunities presented by this technology are immense. One of the opportunities lies with the promise of fungible computing, where you can split the hardware into smaller units but you still have functional items:
Right now, computers are not fungible. With programmable matter, they would be. That same cubic meter of a billion catoms is essentially a network of a billion computers. That’s a lot of computational power – more than enough to organize it into different shapes. And if the computer was separated into sections, the overall computing power would still be the same.
By making “tech” modular in this way the notion of discrete machines for different tasks goes away – you have a generic, all-purpose substance that you can lump together (like clay) to make the things you want.
DARPA are still at it busily inventing the all the science-fictional goodness we expect and deserve. Now they’re going in for programmable matter, of a similar flavour to that found in Fire upon the Deep by Vernor Vinge, Accelerando by Charles Stross, and Dune: The Butlerian Jihad by Brian Herbert and Kevin J. Anderson. The goal of the 
