MIT computer scientists Daniela Rus and Kyle Gilpin are behind this new algorithm, which they are currently testing on 2D grids with "smart pebbles", which are about the size of a cubic centimeter. The key idea here is to get each of these pebbles to communicate with each other and exchange information while using as little processing power as possible. To accomplish that, each of the pebbles are equipped with four magnets, one on each of its sides other than the top and bottom.
These smart pebbles are then placed in a big 2D grid. Part of the grid is missing pebbles - this blank space represents the object the pebbles are charged with recreating. The pebbles then test which magnetic connections are in place and which are missing. Through this process, they can then work out the outline of the missing space and start assembling the shape elsewhere on the grid.
In principle, the smart pebbles could be programmed to recreate the shape at ten times its original size, or to make twenty copies of the original shape. And there's nothing theoretically stopping this from expanding into the third dimension, though admittedly there are some practical concerns - as Rus and Gilpin explain, they could only fit four magnets in each of the pebbles, which meant they could only play around with 2D shapes.
It will take significantly smaller microprocessors for these 2D smart pebbles to become 3D smart sand, but the cool news is that the major theoretical hurdle has been passed - this is now more about figuring out how to solve the engineering challenges. If that can be done, the possibilities of this smart sand are pretty much limitless. For more, check out MIT News.