2004-02-10 Robots - Saving Money, Time and Lives http://conferences.oreillynet.com/cs/et2004/view/e_sess/4844 iRobots make 80% of all mobile robots sold in the US. Robotic vacuum cleaning is a great application: people shouldn't have to wake up in the morning and think "I have to do the vacuuming today". They've moved from $10000 robots to cheaper robots at certain price points: they really needed that discipline and learned it from making toy robots for Hasbro. [interesting that it's the same company that does the toys and then the serious apps.] http://www.irobot.com/consumer/hasbro-irobot.asp [I *so* want a Roomba. Can you get them in the UK? I'm going to buy one. So awesome.] says tom: http://www.buyroomba.com says phil: http://www.letsautomate.com/10911.cfm? also found: http://www.mower-magic.co.uk/ (via perfect.co.uk) one of the problems with infra-red for wall following is stuff like wall *colour*, reflectivity, etc. wicked. wow. and the roomba will "protect itself from doing something catastrophic" and will not fall down the stairs. it also has "escape algorithms". they call them "lobster traps" -- situations in people's homes where it's easy to get into a situation but really hard to get out. they've worked a lot on this. [I like this concept that the environment for robots is like the environment for people but has different affordances, so by looking at the environment at an angle - taking the robot slice - a different picture becomes visible: one with dead-ends and different strange attractors.] [of course, the next step of protecting itself is to protect its battery life by killing you in your sleep so you don't make the house dirty and cause it work.] "Robots are a completely unique class of systems." example being cars aren't just horseless buggies, they've afforded all kinds of stuff. [ok, so we've got systems that diverge and converge (gardens and knots) which have different relaxation times (with an infinite relaxation time these are walls and wormholes; mazes and black holes; corridors and rooms). on the cusp between gardens and knots there are continually unfolding systems: the universe, life - which proceeds in generations/iterations - and electronics: electricity gives us a constructivist cyberspace. switches and relays let systems continually be adapted and unfolded into potential space by people, in a way that mechanical systems can't (you run out of power to drive the cogs: the abstraction layers for adaptive design are too leaky). so robots are another step towards making our virtual worlds into cybernetic systems -- but they're second order but they include a certain amount of recombinance: they can interact with the environment without having human mediation.] ooh, nice app: a Intelligent Mobile Device that maps an unknown space and produces a CAD model you can import into your computer. [imagine a desktop app to help with decorating your house that comes with a little robot to map your house.] bits of innovation, coming soon: "react to prosody" -- the intonation in your voice. "*bad* robot." "take turns in communication" -- conversation. "find power and recharge" this is awesome. home robotics adaptation roadmap: 1. auto appliances: you don't worry about washing your windows or cleaning the floor anymore. [this is such a powerful factor. so far it's all just got more complication. applications to help *more* people do a household budget. you dial your *own* number on the phone instead of the switchboard doing it. robots are like having a telephone operator in your own home. they actually reduce workload instead of making more possible.] 2. no hassle maintenance 3. human assistance and care (like http://www.helpinghandsmonkeys.org but with robots) to perform the task "go get my glasses" you need to solve the vision problem: "vision as an inverse problem." computer graphics: 3d model or scene -> image vision: image -> 3d model or scence (which has inherent ambiguity) [this is the constructivist thing. the robot has to construct a model of reality to operate on, the same as we do (to get around nerve slowness). the interesting thing about cyberspace, scientific models, metaphors, virtual worlds, is that we're externalising our mental models. but we're the only species that does that in a big way (i kind of live in a dog-universe too, but we key on different senses so we don't overlap too much). but robots will have to interface with the human extelligence too, which means there's going to be a big change: we're going to start living in a work with other-species extelligence contributing to our culture. whoa. and it will be different too: Ludo has rules, but the unfolding of rules produces a game. robots have programming, but their interactions with the environment will unfold in new and interesting ways.] [this packbot robot is insane. being thrown through windows then exploring a hotel. rugged, stair-climbing, military carry-able robot. you should be able to control these things with a playstation. perfect interface. i want one to send shopping for groceries. wow, and it's only $45,000] [Rendezvous IM with Matt Jones 10.16 mw: do you reckon you can get a playstation interface to these things? mw: i want to send one to the shop to get bread and milk mj: it's pretty impressive / scary hey? mw: terrifying mw: this + smart dust + 3d printers change warfare. no supply chain problems, easy spying mj: fax the robots into the warzone] [why aren't these guys building the mars probes? bulk manufacturing them for $50,000 a pop and dumping hundreds of them up there?] network centric approach to packbots -- all robots part of the common system. but simultaneously, give the robots more autonomy so you don't need individuals to control each one. [this is insane. we're closer to Spiders than i thought.] [mechwarrior/exoskeletons is such a 1980s vision. no network, no swarming.] Now she's talking about swarms: large >100 numbers of robots, disposable, completely distributed system (no centralised control at all). she's talking about swams of 1000s or 100,000s. great video of the whole swarm in a new building. they found a tiny object in a new building. self-changing, self-collecting, self-organising swarm. and this is a proper demo. # cory asks a question about feral robots. somebody's teaching high school kids to take apart toy dogs and follow volatile chemical gradients. this other company works on the idea of "legible outcomes" -- the outcome of the output has to be human readable. # question on ethical constraints. like, would irobot sell robots to law enforcement to use against protestors? [i'm reminded here of cisco providing a censor firewall to china. ethically unsound.] she's not really answering the question, talking more about safety.