MIT’s shapeshifting 'Kinetic Blocks' can learn from humans

Kinetic Blocks, an invention from MIT's Tangible Media Group, is a grid of pixels that can manipulate objects to create or disassemble structures. Kinetic Blocks can even watch a human build a structure, then replicate that structure on its own.

When it is at rest, the Kinetic Blocks project, created by the Massachusetts Institute of Technology’s Tangible Media Group, doesn’t look like much.

The flat, white sheet, about a foot and a half on a side, is divided into a grid of 900 smaller squares, each of which can raise and lower itself independently. But when it’s being controlled by a computer, the sheet becomes a data manipulation engine, capable of interacting with objects on its surface, assembling them into simple structures, and even watching and learning from human actions.

Kinetic Blocks is a followup to inFORM, a beefed-up version of those pin-screen toys found in offices worldwide. inFORM allowed users to physically interact with objects over the Internet, allowing two people Skyping to shake hands or manipulate a physical object together.

But where inFORM was relatively clumsy, Kinetic Blocks is fast and accurate, with the ability to move, rotate, stack, and twist blocks in an endless variety of shapes.

A video posted by MIT shows Kinetic Blocks’ pixels flickering up and down many times per second to perform tasks such as shuttling a block around the screen’s surface, flipping a cube around all three of its axes, and stacking blocks into small buildings and bridges.

Kinetic Blocks can even perform very small demolitions by nudging assembled structures so they fall over.

Like inFORM, Kinetic Blocks can “see” with a Microsoft Kinect camera and is controlled by a programmable computer. The computer can either store plans for particular structures or learn from watching a human move blocks across the surface of the board.

Special kinematic blocks give the screen additional functions, such as pointing to a particular object or analyzing and replicating a structure.

Current industrial assembly lines require either trained human operators or a series of sophisticated robots to create objects en masse, but Kinetic Blocks suggest a different possibility: A long stretch of 3D pixels capable of taking small components and assembling them into finished products.

Kinetic Blocks’ individual pixels aren’t small enough – yet – to be able to manipulate objects with much precision, but a refined version could dexterously manipulate electronics or other delicate objects.