By Joel Warner
By Michael Roberts
By Alan Prendergast
By Michael Roberts
By Michael Roberts
By Amber Taufen
By Patricia Calhoun
By William Breathes
Gamma Two, which uses the Greek word for the letter G to reference its founders, Gunderson and Gunderson, is dotted with examples of the couple's accomplishments, beyond the two Ph.D.s, two master's and three bachelor's degrees they boast between them. On the wall, near a poster of a drunken robot that reads "A sober robot can't do his job. He depends on your beer!" is a patent for an automated medical-sample freezer Jim designed; its robotic machinery operates at -80 degrees Celsius. And in the corner, there's Kitty, a low-cost hazmat-monitoring robot they developed — built into a blazing pink Barbie Power Wheels car.
Basil is just a few months old ("I went into service on October 15, 2008, at the Gamma Two Facility in Denver, Colorado, USA, Earth," he'll say in an ode to HAL 9000, the creepy supercomputer in 2001: A Space Odyssey). And he's only a hodgepodge of parts from Home Depot and Radio Shack, a thrifty provenance that's allowed the Gundersons to work on him for four years using savings from lucrative past projects, such as the software they developed that played the stock market for hedge fund managers. But simple little Basil could be their most important creation.
"Our motivation is, 'Where is my robot?" explains Jim. "For fifty years, sixty years, they've been promising us our robots. The personal servants who are going to clean our house, walk the dog, do all that kind of stuff. We want Rosie from the Jetsons."
The Gundersons aren't the only ones who dream of a servicebot. As the baby boomers shuffle into their golden years, helpful automatons that allow the elderly to live at home longer could prove to be hugely cost-effective, given the price of nursing homes.
But until now, no one's figured out how to make such robots work, despite the fact that the world already has the hardware to do it. On one side are the amazing bots that stick to simple tasks. An automated car built at Stanford University made headlines in 2005, for instance, by winning a driverless-car competition, successfully navigating a 132-mile off-road course in under seven hours. Then there are the 2.5 million Roombas, the robotic vacuum cleaners that scurry about floors all over the world.
On the other side are the computers with robust artificial intelligence. In 1997, Deep Blue, a chess-playing computer, beat world champion Garry Kasparov and became the best player in the world. And Jim, for his dissertation, developed a computer program that could deliberate. When given a task, the software cycled through every action it knew how to do, picked the most suitable ones and figured out in what order to do them.
But when such AI is installed on robots tasked with getting stuff done in the real world, things get messy. Deep Blue, for example, knows how to move a knight piece on a chess board better than any human, but it doesn't actually know what a knight is — or, for that matter, where to find such a piece in a cluttered cabinet full of board games.
"We want robots that we can tell what to do and they'll figure out how do to it, and if something goes wrong, they'll figure out how to fix it or come back and ask for help," says Jim. "Complications, no problem. Changes with the world, no problem."
So far, that goal's been a pipe dream.
That is, the Gundersons say, until Basil came along.
There's a problem. Not a big problem, but a problem. It's just two weeks before the Cafe Sci meeting at the Wynkoop, and Basil is completely lost. He's still moving around, rumbling from one point to the next when he's told to, but he never ends up at the place in the lab where he thinks he is. The Gundersons know his location is off because they've measured out the distances in their lab to the millimeter.
So they stay hunched over their computers, tinkering with code before plugging Basil in and downloading minute revisions. But the glitch won't go away; every time they seem to iron out one wrinkle, another one pops up. It's one of the many headaches of nuts-and-bolts robotics — while purely software-based artificial-intelligence programmers can run thousands of software tests to root out bugs almost instantaneously, fixing Basil means endless hours of slow, careful trials. After the latest round of fiddling, Basil seems more perplexed than ever. When directed to move across the room, he turns a bit and shuffles forward a few inches — then stops. "What would you like me to do?" he asks, believing he's somehow teleported himself clear across the workspace.
"That's even weirder," remarks Louise, sipping a cup of smoky oolong tea — the couple's umpteenth of the day. "I must have really broken something." Over the lab speakers, a chipper Christmas tune plays — one of many the "Personal DJ" software the two designed has been spinning lately. The AI program has learned enough about the couple's musical tastes to mix things up, but ever since Jim, in a bout of seasonal cheer, uploaded a bunch of holidays tunes, it just can't seem to help itself.