Personal fabbers on the horizon
One component of Professor Hod Lipson's research—which explores if computers can accelerate and augment human invention—bristles with monumental implications. Imagine owning a machine, about the size of your desktop printer that, at the touch of a button, will churn out three dimensional, completely functional products such as cell phones, shoes, even food and prosthetic limbs. Lipson envisions a world where people don't sell or distribute objects to each other, but exchange blueprints of things that they “print out” for themselves on their personal “fabbers”. With prototypes already being tested, the era of “personal fabrication” may not be far off.
“This can really change the way that we ship, buy, make, and stock things,” says Lipson, who is an assistant professor in Cornell's Department of Mechanical and Aerospace Engineering and affiliated with CIS. “But most importantly, it will allow the exploration of a new design space, and greater freedom—both for humans and for computers—to create new things.”
This momentous goal is Lipson's mission: to teach computers to design and manufacture their own offspring.
“Engineering and science are moving into complexities and dimensions where people have little or no intuitions. One way out of this conundrum is to design machines that can design for us.” writes Lipson in a recent article on “evolutionary robotics and open-ended design automation.”
The first halting steps were taken in 2000, when Lipson (then at Brandeis University) and Jordan B. Pollack reported the successful fabrication of a robot—nothing more than a self-propelled bundle of trusses, motors, and wires—that was wholly designed and fabricated by another machine. The “mother” computer used what is called an “evolutionary algorithm” to search the range of possible robotic designs and test them against real-world constraints, basing its products on the designs that competed most successfully. Designs that “survived” best included a model that pushed itself along like mechanical inchworm.
“That was back in 2000—we have lots of other examples, now, in a variety of disciplines from finite state machines to gene regulation networks.” says Lipson, who is currently using the mechanism of coevolution—machines competing against machines, as in biological predators vs. prey—to accelerate computer-generated innovation. Speaking in The Washington Post, famed MIT robotics expert Rodney Brooks hailed Lipson and Pollack's accomplishment as “...a long-awaited and necessary step toward the ultimate dream of self-evolving machines.” Recent work on physical self -replication takes further steps toward physically adaptive robotics.
Not surprisingly, support for Lipson's work has come from agencies such as NASA, which could make good use of a general-purpose fabber on deep space missions, and NIH, which could make use of new algorithms for inferring complex biological systems. Opportunities exist now for undergraduates to become involved in technologies that can, at the very least, revolutionize our world as thoroughly as the computer itself. At the most, they may well hasten the dawn of a new evolutionary age.