Wednesday, November 03, 2004
Robot Capitals and more RoboNexus videos...
Here are the locations of 3 videos taken at RoboNexus 2004
Mine - http://www.plyojump.com/robonexus.html
Mobile Robotics/VIA -
RoboNexus (several videos, includes many by TV news programs) - http://www.robonexus.com/video.htm
Also, a nice article on how Osaka is angling to be the "robot capital" of Japan, and is hosting a huge tradeshow and Robocup in 2005.
http://www.asahi.com/english/business/TKY200411020099.html
Here are the locations of 3 videos taken at RoboNexus 2004
Mine - http://www.plyojump.com/robonexus.html
Mobile Robotics/VIA -
RoboNexus (several videos, includes many by TV news programs) - http://www.robonexus.com/video.htm
Also, a nice article on how Osaka is angling to be the "robot capital" of Japan, and is hosting a huge tradeshow and Robocup in 2005.
http://www.asahi.com/english/business/TKY200411020099.html
Tuesday, November 02, 2004
Robotics motto: Free Your Body
One of my favorite old TV shows, is The Prisoner, a classic 1960s "Awakening" or "counterculture" statement against the faceless, dehumanizing power of the State. One of the key features of stories from this time period is the emphasis on navel-staring - the focus is on what goes on inside - one's "head trip" rather than external events.
A good example: today, in a violent movie fight, the hero receives a blow to the forehead. In today's "external" cinema, the camera zooms in to show the damage to the hero. In contrast, in The Prisoner and other stories from its era, a blow to the head is simply an excuse to show us the "head trip" experienced by the hero as he lapses into unconsciousness. Which such a focus on "the land inside your mind" (Moody Blues) it's not surprising that the message from The Prisoner is this: "...free your mind..."
Today, robotics needs to do the opposite. The first serious efforts at non-industrial robotics date from the late 1960s, the era of The Prisoner - but for many practicioners, time has stopped. Despite the widespread refutation of "classic AI" disembodied mind from that Hal 9000 era, many robotics workers, along with the general public, assume that the problem with robots is that their minds are not good enough. A recent article on ZDNet contained the following quote from Joe Engelburger, the "father" of industrial robots:
"You can get a robot to speak with you, but you can't get it to talk Spinoza with you"
Leaving aside the fact that few college students today could tell us any more about Spinoza than a robot, this "free your mind" vision of robotics is off base. Who cares? Spinoza is interesting, but I doubt he is necessary to make real robots. I must say I wonder if a generational effect is at work here - Engelburger's accomplishments hail from the same "free your mind" era of the 1950s and 1960s.
Even Hans Moravec, co-founder of SEEGRID and author of several mass-market books on robots, phrases the problem in terms of mind - a "guppy brain" is all we need for useful robots. True, but that isn't really the trouble with robots. We can make guppy brains, more or less.
The real problem, ignored by many until recently, has been the wretched bodies made for robots. Created out of cast-off parts from other industries, nearly every robot made is a junk-bot. A engineering approach caused by cost issues and minimizing parts has led to nearly all robots making do with a tiny bit of sensor data. Apparently, clever processing can compensate for the limited data from say, and IR sensor, and multiple sensors are a disadvantage since they lead to more complex decisions being required with increasing ambiguity.
Um, nonsense. If this were the case, we would expect living things to attempt to minimize their biological sensors - after all, if more sensors simply lead to wasteful extra processing, evolution should try to minimize them. Brain tissue is very energy intensive, as are sensory nerves, and if it were possible to reduce their number, they would be reduced.
But in virtually all animal phyla, sensory tissue reigns supreme. Animals with almost no brain are nevertheless provided with an incredible density of sensors which detect a variety of stimuli - temperature, pressure, light, chemicals.
Even bacteria, simple single cells without a nucleus have these same temperature, pressure, light, chemical sensors - hundreds of them per cell. The Sea Slug, a classic neurobiolgy animal because its "brain" consists of a few dozen giant neurons, nevertheless has millions of sensory nerves throughout its body. The primacy of sensors over the "mind" is evident.
Why is so much sensation necessary? The key to moving in the real world is not mind, but body. A properly sensitive body, even with simple "brains" is far superior to a nearly blind and deaf robot with a huge processor behind it.
In addition to fine-grained sensation, fine-grained motion is key to robots. Animals could move their joints with a few dozen muscles. Instead, there are over 600 muscles in the human body. Simple animals (like the infamous Sea Slug) have even greater degrees of freedom - they can reshape their plastic, well, slug-like body at will into arbitrary forms. Once again, if there were an advantage to a simple body, evolution would take that course. Instead, an extremely fine-grained control of the body, coupled with a huge number of "actuators" is the rule for living things.
It is tempting, in the "junk-bot" dark ages of robots today, to build a simple body and work on the mind. It is easily to manipulate software, hard work to make hardware. But the major unfinished task today is hardware. We need complex, fine-grained robot bodies - the more the better - preferrably built of of redundant simple components. The issues of control of extremely complex sensor and motor networks has not been a major topic in robotics, partly because robot bodies with these features do not exist. But the first groups to replicate biological complexity in robot sensory and motor systems will "free the body" - the mind can follow along later.
In this respect, some of the major changes will be the creation of massively "multicore" processors specifically for robots. A PC can benefit from a dual-processor configuration (one for the OS, one for the application) but more cores will have limited effects. In contrast, a 1000-core processor, if it can be built, should be ideal for implementing the parallel processing required for hierarchical, massive sensor and motor networks.
So...mechanical engineers - forget the mind...free your body.
One of my favorite old TV shows, is The Prisoner, a classic 1960s "Awakening" or "counterculture" statement against the faceless, dehumanizing power of the State. One of the key features of stories from this time period is the emphasis on navel-staring - the focus is on what goes on inside - one's "head trip" rather than external events.
A good example: today, in a violent movie fight, the hero receives a blow to the forehead. In today's "external" cinema, the camera zooms in to show the damage to the hero. In contrast, in The Prisoner and other stories from its era, a blow to the head is simply an excuse to show us the "head trip" experienced by the hero as he lapses into unconsciousness. Which such a focus on "the land inside your mind" (Moody Blues) it's not surprising that the message from The Prisoner is this: "...free your mind..."
Today, robotics needs to do the opposite. The first serious efforts at non-industrial robotics date from the late 1960s, the era of The Prisoner - but for many practicioners, time has stopped. Despite the widespread refutation of "classic AI" disembodied mind from that Hal 9000 era, many robotics workers, along with the general public, assume that the problem with robots is that their minds are not good enough. A recent article on ZDNet contained the following quote from Joe Engelburger, the "father" of industrial robots:
"You can get a robot to speak with you, but you can't get it to talk Spinoza with you"
Leaving aside the fact that few college students today could tell us any more about Spinoza than a robot, this "free your mind" vision of robotics is off base. Who cares? Spinoza is interesting, but I doubt he is necessary to make real robots. I must say I wonder if a generational effect is at work here - Engelburger's accomplishments hail from the same "free your mind" era of the 1950s and 1960s.
Even Hans Moravec, co-founder of SEEGRID and author of several mass-market books on robots, phrases the problem in terms of mind - a "guppy brain" is all we need for useful robots. True, but that isn't really the trouble with robots. We can make guppy brains, more or less.
The real problem, ignored by many until recently, has been the wretched bodies made for robots. Created out of cast-off parts from other industries, nearly every robot made is a junk-bot. A engineering approach caused by cost issues and minimizing parts has led to nearly all robots making do with a tiny bit of sensor data. Apparently, clever processing can compensate for the limited data from say, and IR sensor, and multiple sensors are a disadvantage since they lead to more complex decisions being required with increasing ambiguity.
Um, nonsense. If this were the case, we would expect living things to attempt to minimize their biological sensors - after all, if more sensors simply lead to wasteful extra processing, evolution should try to minimize them. Brain tissue is very energy intensive, as are sensory nerves, and if it were possible to reduce their number, they would be reduced.
But in virtually all animal phyla, sensory tissue reigns supreme. Animals with almost no brain are nevertheless provided with an incredible density of sensors which detect a variety of stimuli - temperature, pressure, light, chemicals.
Even bacteria, simple single cells without a nucleus have these same temperature, pressure, light, chemical sensors - hundreds of them per cell. The Sea Slug, a classic neurobiolgy animal because its "brain" consists of a few dozen giant neurons, nevertheless has millions of sensory nerves throughout its body. The primacy of sensors over the "mind" is evident.
Why is so much sensation necessary? The key to moving in the real world is not mind, but body. A properly sensitive body, even with simple "brains" is far superior to a nearly blind and deaf robot with a huge processor behind it.
In addition to fine-grained sensation, fine-grained motion is key to robots. Animals could move their joints with a few dozen muscles. Instead, there are over 600 muscles in the human body. Simple animals (like the infamous Sea Slug) have even greater degrees of freedom - they can reshape their plastic, well, slug-like body at will into arbitrary forms. Once again, if there were an advantage to a simple body, evolution would take that course. Instead, an extremely fine-grained control of the body, coupled with a huge number of "actuators" is the rule for living things.
It is tempting, in the "junk-bot" dark ages of robots today, to build a simple body and work on the mind. It is easily to manipulate software, hard work to make hardware. But the major unfinished task today is hardware. We need complex, fine-grained robot bodies - the more the better - preferrably built of of redundant simple components. The issues of control of extremely complex sensor and motor networks has not been a major topic in robotics, partly because robot bodies with these features do not exist. But the first groups to replicate biological complexity in robot sensory and motor systems will "free the body" - the mind can follow along later.
In this respect, some of the major changes will be the creation of massively "multicore" processors specifically for robots. A PC can benefit from a dual-processor configuration (one for the OS, one for the application) but more cores will have limited effects. In contrast, a 1000-core processor, if it can be built, should be ideal for implementing the parallel processing required for hierarchical, massive sensor and motor networks.
So...mechanical engineers - forget the mind...free your body.
