PAY ATTENTION TO DETAILS
DANNY HILLIS'S PAINT CAN ROBOT
1956ā
Artist interpretation of Danny Hillis's robot made from paint cans. (See endnote.)
Each year, the MIT computer science department picks a few theses out of its graduate student harvest for publication. In 1985, Danny Hillis's landmark development in computer hardware was one of them. His Connection Machine was revolutionary. It was a ācomputing engineā in which processing was accomplished concurrently through a massive number of simple processors, with a resulting substantial increase in speed and efficiency. Simply put, processing in the Connection Machine occurs in parallel, as opposed to the sequential processing of a linear processor. In a basic scenario, one instruction is executed at a time in linear processing, whereas in parallel processing, numerous instructions are executed simultaneously, which saves time. If a computer is a restaurant and the processor is the kitchen, then Hillis's parallel computer is like having a team of cooks to prepare the meals, instead of a single chef; the food gets served a lot faster.
Despite its success, Hillis says the parallel computer was ahead of its time. While Hillis foresaw a market for it from the outset and set up the company Thinking Machines to sell it, it did not do as well as expected, and the company eventually dissolved. He says that there was so much emphasis and investment in the personal computer back then that the market was unwilling to change the software. It took more than 20 years, but parallel processors now predominate modern personal computer architecture (i.e., dual-core or quad-core processors), and they have been used for years in āsupercomputersā for a wide variety of research and commercial applications.
In contrast, the raid (redundant array of independent disks) hard-drive system Hillis developed for the Connection Machine, which uses multiple hard drives simultaneously, was a big commercial success from the outset. Hillis, however, almost did not patent it. He thought that such a clunky system, which uses mechanical motors, was surely technology on its way out. Boy was he wrong. It just goes to show that inventors ālive in the future.ā
Such forward thinking is evident in Hillis's other activities. In fact, he's a founding member of the Long Now Foundation, which promotes long-term thinking. The organization is bent on adjusting our idea of time and stretching it well beyond our own puny experience to the vast, and more realistic, timescale of the Earth. The concept was Hillis's:
The first prototype of a 10,000-year clock-of-the-long-now, completed by Hillis in 01999, is housed in the Museum of Science, London. Another clock is in progress that measures the same time frame but works on the basis of a sun and planet model ā a modern astrolabe. The workings of the clock are technologically unique; it has a precision equal to one day in 20,000 years, and it self-corrects daily according to the noon sun. Hillis has several patents from this work. The final design for the clock will be very large. It will eventually rest on a mountain in Nevada with a white limestone face, which has been purchased by the foundation for this purpose.
Hillis has also turned his attention to the Internet, with the creation of a new project, www.freebase.com. The website is described as a āsocial database about the things you know and love,ā and it was publicly launched in beta phase in 2007. This public database is to Wikipedia what the Connection Machine is to the von Neumann computer. While Wikipedia's database has a hierarchical structure, with each bit of information having its own place as a subset of a bigger category, Freebase is flexible. Each collection of information is thought of as a hub, with each hub connected to many others. Bits of information can belong to however many hubs people decide they do. To this Internet grazer, who is a gentle herbivore, so to speak, rather than an autotroph or carnivore of Web technology, Freebase seems to be powerful stuff. It has a graphic feel to it and gives the impression of a leap forward in how knowledge is stored and used. Freebase encourages users to create novel applications that use the database and provides some tools to do so. With Freebase, Hillis is taking his ideas of a brain machine one step further ā he seems to be aiming for a global brain with this invention that is analogous to the prefrontal lobe or hippocampus or perhaps an altogether unique kind of neural network. Hillis shared something along this line of thought with the web magazine Edge:
There are other works in Hillis's oeuvre, and he holds 50 United States patents to date, ranging from simple devices like a laptop computer desk stand to more sophisticated lens systems, sensors, and an apparatus to maintain eye contact during teleconferencing. He has two patents for methods of masking speech, with applications for anyone who wants to speak on the phone with privacy. There's also an atmospheric spectrometer for use by planetary satellites, a method for locating and identifying underground structures, a control wand, a robot for military use on the field, a 360-degree awareness system for drivers in a car (instead of just rear and side-view mirrors), and more.
Many of Hillis's inventions, including the parallel computer, have been part of a broad interest in artificial intelligence (AI). In the introductory pages of his PhD thesis, The Connection Machine, he compares the human brain to the computer, pointing out the ostensibly disparate attributes of the brain, its relatively slow operating speed coupled with its ability to outthink a machine with ease:
The excitement and frustration in Hillis's words about the potential and challenges of AI is almost tangible. His fascination was nurtured in tandem with efforts made in inventing robotics. Hillis's first robot was a spectacular piece of engineering. Standing several feet tall, the humanoid had a head made from an empty 1-gallon paint can sitting on a trash-can body. Its four limbs were made from 1-quart paint cans. The Christmas treeābulb eyes lit up, and it had a speaker for a mouth, wired for sound. Its arms moved via the use of a rotisserie motor, which was rescued from a discarded barbecue. Hillis built this robot with the explicit purpose of having a companion to walk with him, or at least bring him orange juice in the morning, and he was exceedingly disappointed when the invention failed. Like all inventors, Hillis had a vision, and he built that vision. But as is common in the inventor's process, he had neglected to visualize a key detail required to make the invention functional. In this case, he'd forgotten to give his robot a brain. Danny Hillis was only nine years old at the time, so his spectacular failure is probably forgivable. Besides, in the broader context, this particular paint-can robot served a very useful purpose. For in that pivotal moment when the robot did not do his bidding, Danny realized the importance of the āmachine brain,ā and his lifelong interest in computers began. We have this paint-can robot to thank for that interest, and the substantial benefits to society it later produced.
At a much later date, Hillis did make a working robot, fulfilling his childhood dream ā and, in a sense, repeating his childhood failure. While vice president of research and development for Walt Disney Imagineering from 1996 to 2000, Hillis wanted to create a unique attraction for theme park visitors, something sensational that would provide thrills and entertainment in keeping with Disney's philosophy. Hillis wanted to build a robot dinosaur that wandered around the park, interacting with the public as it did so. The executives were not convinced this was such a good idea. Compared to riding a roller coaster, a wandering metal dinosaur sounded a little low on the thrills side of the equation. But Hillis was given a green light to build a prototype.
When the day arrived to reveal his masterpiece machine, Hillis knew he had a challenge ahead of him. He needed to prove to his colleagues that being approached by a metal dinosaur was indeed a thrill worth paying for, one that would evoke wonder and awe. Hillis pl...