Social Robotics, Scmocial Robotics: Feral Robotics and Some Other Quacking, Shaking, Bubbling (what would the opposite of feral be?) Robots Natalie Jeremijenko http://conferences.oreillynet.com/cs/et2005/view/e_sess/6310 At the O'Reilly Emerging Technology Conference San Diego, California, 16 March 2005 Impressionistic transcript by Cory Doctorow doctorow@craphound.com -- I work on examining and experimenting with the cultural opportunities provided by technology. The emerging field of collective semi-autonomous robotic platforms assumes the way that we get social is by assuming that social means "more than one." But this doesn't get to distributed interpretation... Here's my exploration of social robotics as more than a handful of robotic platforms, about more than robots, people too. Feral Robotic Dogs: It's a website. Everything reduces to a website. A couple years old, dates back to the launch of the Sony Aibo. One in a series of interactive toys that express behaviors programmed in our labs -- they're fun and interesting and sci-fi-ey. But what do you learn from them? You learn construction from construction toys, monopolization from Monopoly. What do you learn from interactive toys? Interaction? These toy dogs out of the box beg for bones or sing the national anthem. I became interested in this when someone said to me that a robot dog would make a good pet for me -- what does that say about my capacity to care for living things? What might we learn from these things? What do we need to learn from these things. Here's the website (xdesign.ucsd.edu/feralrobots) with instructions for upgrading the raison d'tre of your robot dogs. Warning label: OUT THERE IN HAPPY FAMILY HOMES IN THE OFFICE OF CORPORATE EXECS, IN TOY STORES THROUGHOUT THE GLOBE IS AN ARMY OF ROBOTIC DOGS. THESE REMI-AUTONOMOUS ROBOT CREATURES, THOUGH CURRENTLY PROGRAMMED TO PERFORM INANE OR ENTERTAINING TASKS, ARE ACTUALLY FULLY MOTILE AND AWAITING FURTHER INSTRUCTIONS. People who get toy robot dogs don't play for them for long. We take in a lot of abandoned robotic dogs. The entertainment lasts 15 minutes and then they're boring. The site has: * Instructions for mechanically altering the dog to equip them for all-terrain performance by upgrading the drive-train and amputating the legs, widening the wheels, etc * Instructions for putting in a new nose that senses environmental toxins * Instructions for installing a new brain that follows concentration gradients in toxins, so that the dog can sniff them out All resources and instructions are available on the website. We've run this seven times in classes with groups of kids. Why do we do this? 1. Exploits the economies of scale of corporate toy production -- by far the cheapest robotics platforms in the field, as little as $2 at Kay-Bee Toys. These things cap out at $40 -- the Aibos aren't inexpensive enough to support and sustain an open source robotics platform. The largest market for Aibos in engineering schools, where the academic discount brings the $2700 price-tag down to $2500. You buy a bunch of them and can compete in a Sony-sponsored robot-soccer event. What do you learn from this? 2. These dogs are everywhere -- there's a team at Cornell, SF, Dublin, London, Brisbane, UCSD -- there's an online distributed community for people interested in transforming toys into activist tools. These are designed to be released at a site of community interest. There's a park in Miami that is a residential area that used to be a naval base. It's toxified with volatile organic compounds, and the developers claim that it's been sealed with topsoil. The developers are building their school on the toxic site -- there are 160 schools built on or near toxic waste superfund sites. Schools are poor and toxic waste sites are cheap. There are many interested people, but the EPA reports are often 15 years out of date. This information is often not public. Releasing packs of robot dogs creates a mediagenic event -- because the dogs appear to be sniffing something out, they display the information through their movement. A four-year-old and a grandmother can understand what these dogs are doing and what it means to them. [[Shows video of 15 year old Bronx kids who'd never done engineering at all]] Kids were freaked out by being able to take apart electronic toys and screw with them. Bronx's Starlight Park has been a residential site for 50 years. The area has the highest respiratory death rate in the country. The sites have been known to be toxified. We invite in the media and local regulators. We did this today at the MIssion Bay landfill where military contractors dumped several thousand barrels of toxic waste in an unlined landfill in the landfill adjacent to Sea World, where people swim and wind-surf. It's contaminated with leech from the dump. We invited in the local government, including a mayoral candidate. What happens when you change who has the evidence? You change the structure of participation between the lay audience and the expert audience. You change who is involved in the political process. Those 15 year olds in the Bronx are at all the hearings, they're on TV, they're the experts. They ARE. They know more about it than I do. There are designs that I had nothing to do with -- more than 20 on the website. It changes who gets to have an opinion, who gets to participate. The question of safe or not safe, clean or not clean is complex -- it requires the distributed interpretation of many people. Greg Ellin: How do the dogs report back on the data to a central repository? Natalie: These robots privilege the real-time sensory interpretation of the people on the ground. With a toy robot, you put the cameras where the eyes should be to provide the illusion of life and agency. With a feral robot, the camera goes on the ass, to capture the human sense-making interpretations as they follow the dogs. We had a simple system -- the dogs whistled at each other when they found toxins above a certain threshold. They'd converge on the whistler and bark the national anthem, etc. The cost of coordinated action is fairly high, so it's often dropped. The coordination is "through the ground" because they're responding to the same system. These things have 8k of onboard memory and we write the moving averages of the sensory output and recover it later. Feral robot dogs are open source robots. The adaptations and results are published. There's a growing community of people interested in using these toys not for interaction, but for exploring complex environmental issues of contaminants created by the production of consumer electronics (Silicon Valley has more superfund sites than any comparably sized region in America). IT isn't clean -- it's a problem for the young kids. -- Robot goose: Here we take outdoor technology and bring it into urban contexts. We take hunting tech and bring it into the city. How do we re-script how we interact with nature through our technologies? This is the first OOZ project (Zoo backwards without cages). Here's a robotic goose that can drive around and interact with other geese. The controller is a school desk -- when you lean forward, the goose goes forward, when you lean back the goose leans back. You control the goose and you use it to play with, terrorize and chase real geese. If you've got a problem with geese you're already interacting with them. Geese have a lot of chutzpah. Not only can you play with the geese, you can talk to them, with prerecorded goose words. The geese haven't attacked it yet, but they aggressively hiss and talk to it. Every time you utter a prerecorded sound to a goose, the robot samples the response and uploads it to a database. The distributed interpretation of many people trying to interact with the geese, thinking about what they mean is the best way to learn goose. Language is an interactive system learned from interpretation and shared meaning. We won't learn it from experts. The real geese are usually terrified of it. Ducks tolerate it. We gave this to people in Echo Lake -- they hadn't even realized that they had geese. The robot goose without real geese is like a Nintendo cartridge without a Nintendo. The goose makes you pay attention to your local geese. The locals now watch their geese avidly, worry that the babies will be harmed by the pollution. -- Ever wanted to talk to fish? Ever felt guilty looking them in the eye on your plate? They're interesting for their social behavior. Like people, they have complex social systems. This device has an architecture of reciprocity. The robot goose gives you all the advantages -- you can see them, but they can't see you. You can chase them but they can't chase you, etc. This is a plank that sits on the water with one degree of freedom, it twists with the flow. There's a light -- when something passes through the hole toward the top, the light goes on on the top and people know a fish is coming. When something passes through the hole toward the bottom, the light goes on and the fish know food is coming. This was designed for the Hudson River. People are scared of feeding and interacting with animals: DO NOT FEED THE ANIMALS DO NOT TAP ON THE GLASS. The signs are as ubiquitous as the desire to interact. Why not feed the animals? Is human food not good enough for them? Will we create a dependency? Interfere with their natural instincts -- crap. We put freeways through their habitats and change the climate -- who cares about changing their instincts? We use edible lures that both fish and people can eat. They have PCB-shielding agents in them. you small actions can accumulate to a remedial effect. There's lots of PCBs in New York -- you can't eat the fish. It's the largest superfund site in the country. Here's how you can do something good and constructive.