Issue link: http://cp.revolio.com/i/393
#1 Fall 2008 11 AI of funding. Japanese companies have been the pioneers here but their enthusiasm has flagged in recent years, with Sony dropping its Qrio project and Honda's Asimo robot remaining, basically, a skunkworks project. e robotics industry as a whole is argu- ably flourishing better than ever, but there is a huge gap between Roombas, industrial robot arms, and their ilk, and mobile hu- manoid robots with the capability for com- plex interactions in the physical and social world. Open source humanoid robots have been proposed before, e.g. PINO created by Japanese scientists and launched in 2001. ese earlier projects were technically solid but didn't really take off in the community. However, I'm guardedly optimistic that the iCub may meet a better fate. Early results look promising – for instance, a nifty vid- eo of iCub drumming (see resource link). (OK, it's no Max Roach yet, but what we do have here is coordination of hands, feet, and hearing – sensorimotor integration – which is a powerful first step toward real embodied intelligence.) Of course, demos are demos, not ro- bust technologies, and making a demo of a robot playing the drums is no big trick, given modern engineering technology. But if you dig a little deeper, you find that the technical ideas underlying the iCub seem extremely solid, and it's clear that the ar- chitecture is capable of a lot more than just the handful of tricks demonstrated to date. Its fingers and arms have an impres- sive number of degrees of freedom: a choice made because the designers favor cognitive theories, implying that advanced human cognition largely arises out of the interac- tion between perception and action in the manipulation of objects. iCub itself is just a platform and it doesn't solve all the problems of robotics, by any means. e iCub team has so far fo- cused on low-level perception, action, and coordination, without plunging much into the depths of communication, learning, ab- stract reasoning, and so forth. But they are collaborating with others that have exper- tise in areas such as language learning. And the beauty of the open source approach is that it's relatively straightforward for others with AI ideas and technical chops to extend their work. Building an iCub of one's own is not free, nor trivial, but it's a damn sight easier than designing your own humanoid from scratch... and more possible than get- ting your hands on Qrio or Asimo, which have not been publicly released. And unlike Sony's Aibo, the robotic dog who has be- come a staple of academic AI research -- if one finds aspects of the hardware platform inadequate, one can always modify it, since the specs are completely open. Different researchers are bound to take the iCub in radically different directions. For instance, while I'm an AI guy rather than a robot- ics researcher, reading about iCub has in- spired me to think a bit about how it might be integrated with various open source AI software platforms, robot simulators. and virtual worlds. Will open source do for humanoid ro- botics what it's done for Web browsers and bioinformatics? It's too soon to say for sure, but there's reason to hope. Ben Goertzel is the CEO of AI companies Novamente and Biomind, a math Ph.D., writer, philosopher, musician, and all-around futurist maniac. Making a demo of a robot playing the drums is no big trick, given modern engineering technology. Resources iCub www.robotcub.org PINO www.symbio.jst.go.jp/PINO iCub Drumming www.robotcub.org/index.php/robotcub/content/ download/1135/3982/file/icubFullDrumming3. wmv Open source AI software platforms Robot simulators www.goertzel.org/blog/2008/05/open-source- robots-robot-simulators.html Strong AI en.wikipedia.org/wiki/Strong_AI Ray Kurzweil www.kurzweilai.net The Singularity Institute for Artificial Intelligence www.singinst.org