Tuesday, September 07, 2004
One-task robots lead the way to robots that jump
PC World recently had an article on the rise of the robots. Writing in the technology->future technology area of the online magazine, writer Eric Butterfield looked at two aspects of the growing industry: single-purpose mobile robots like vacuum cleaners and lawn mowers, and the more general-purpose humanoid robots being created (mostly in Japan) for healthcare.
To me, the most fascinating aspect of this article (and similar ones that are popping up all over the PC world) is the distancing. The reporter describes robots as if his area of expertise, personal computing technology, is as distant from robots as it is from container ships or machine tools. Rather than reporting on robots as the "next thing" for PC power users, the robot industry is described like the Space Shuttle - high tech, but not directly related to personal computing.
This is true despite the fact that robots use the same hardware as PCs - RAM memory, hard/flash drives, cpus, mainboards, buses, USB ports, wireless technology and more.
Why don't these tech gurus have more connection to robotics? Why can't they tell us the details of the processing speed, dataports, or mainboards used in robotic creations? Why don't they have their own personal robot crawling around the room?
What is happening here is that a new technology is being born of the old. Using the same parts, robotics is creating machines that are out of the realm of the PC tech guru or power user. You don't see PC gurus reporting on the internal guts of these systems, despite the fact that it is similar to the high-end gamer's delight PC on their desk. For example, the Honda Asimo uses 4 PowerPC processors with the Wind River operating system - a kind of real-time Unix suited to device control. Technically, you can log into Asimo just like you were a linux geek. But tech gurus steeped in Linux don't report on this - it is essentially out of their depth, and happening outside their cherished industry.
The reason is that robots, despite using the same mainboards and memory chips, as PCs, are fundamentally different devices. With charity, we can define a PC as an attempt to build an electronic lump of clay, which can be configured by users and programmers to particular tasks. Take a drawing program like Adobe Illustrator. Drawing programs are largely passive, and don't have any understanding of what is being drawn in their dataspace by users. Their output does nothing to benefit them or the PC they are running on.
In contrast, robotic software is active, and does try to understand its input instead of simply processing it according to arbitrary rules. The output of the software has a direct impact on the performance and survival of the robot in the real world - a software error could cause the robot to be squished by a car or lose power if it fails to find the wall socket.
There is no passivity here - robotic software must "push back" on sensory input, whereas PC software must passively adjust to user input.
This difference explains the differences in perception of robots and PCs. PCs are the ultimate slave - trying to do exactly what the user wants with no "personal" input. This makes PCs very flexible but utterly dumb - since they don't try to understand the task they are performing, they screw up left and right, and must be brought back by the user. Even with a novice at the keyboard, the PC is the idiot in human-computer interaction, mostly because it has no idea of what is going on.
In contrast, a robot is not quite a slave, despite the popular image of them as mechanical servants. To qualify as a mobile robot, the device must do a lot of processing that does not consist of orders from a human - primarily sensory data roughly equivalent to human senses. The order from a human is an occasional thing in the overall processing.
One might give an order for a robot to turn left. This will be a small amount of the total processing, which will be processing environmental data so that the turn can be executed. The order is not arbitrary to the robot - both human user and robot attempt to understand the real-world environment in a way allowing a turn left.
The profound difference between PCs and robots makes the typical tech guru blind. Despite the fact that their desk hardware could make a credible robot a la PC robots like White Box Robotics, they know little more than a non-tech person about the industry. The articles don't contain tech details, but instead contain commentary by robotics professionals on the future of their field. Someone who puts in fence posts for a living could collect the same information with the same competence. The PC tech guru - despite all their hardware and software sophistication - is on the outside, looking in.
There's a precedent for this. Imagine it is the year 1978 and you're reading Popular Electronics magazine. An occasional article begins popping up about homebuilt personal computers, and attempts to create useful mass-market PCs by startup businesses. Despite the electronics orientation of the magazine, the tech writer doesn't really understand PCs any more than you do - they are reporting on an areas outside their realm. They don't have a homebrew computer or even an Apple I on their desk. This is despite the fact that the homebrew PCs use the same components found in many of the electronics projects described by the magazine.
It is interesting to watch the future of PC-based publications and gurus unfold with the rise of the robots. Will tech gurus continue to watch the rise of the robots as we might watch a report on Mars exploration? Will new robotics gurus such as those being showcased on Everyday Robots become the new tech luminaries? Will the PC industry, along with its trade publications, celebrities, stock market darlings, tech gods and industry centers, be sidelined as the robotics revolution moves forward? The gap between PC and robotic ideas - despite the common hardware and software - argues that the robotics revolution will happen outside of the personal computer world.
PC World recently had an article on the rise of the robots. Writing in the technology->future technology area of the online magazine, writer Eric Butterfield looked at two aspects of the growing industry: single-purpose mobile robots like vacuum cleaners and lawn mowers, and the more general-purpose humanoid robots being created (mostly in Japan) for healthcare.
To me, the most fascinating aspect of this article (and similar ones that are popping up all over the PC world) is the distancing. The reporter describes robots as if his area of expertise, personal computing technology, is as distant from robots as it is from container ships or machine tools. Rather than reporting on robots as the "next thing" for PC power users, the robot industry is described like the Space Shuttle - high tech, but not directly related to personal computing.
This is true despite the fact that robots use the same hardware as PCs - RAM memory, hard/flash drives, cpus, mainboards, buses, USB ports, wireless technology and more.
Why don't these tech gurus have more connection to robotics? Why can't they tell us the details of the processing speed, dataports, or mainboards used in robotic creations? Why don't they have their own personal robot crawling around the room?
What is happening here is that a new technology is being born of the old. Using the same parts, robotics is creating machines that are out of the realm of the PC tech guru or power user. You don't see PC gurus reporting on the internal guts of these systems, despite the fact that it is similar to the high-end gamer's delight PC on their desk. For example, the Honda Asimo uses 4 PowerPC processors with the Wind River operating system - a kind of real-time Unix suited to device control. Technically, you can log into Asimo just like you were a linux geek. But tech gurus steeped in Linux don't report on this - it is essentially out of their depth, and happening outside their cherished industry.
The reason is that robots, despite using the same mainboards and memory chips, as PCs, are fundamentally different devices. With charity, we can define a PC as an attempt to build an electronic lump of clay, which can be configured by users and programmers to particular tasks. Take a drawing program like Adobe Illustrator. Drawing programs are largely passive, and don't have any understanding of what is being drawn in their dataspace by users. Their output does nothing to benefit them or the PC they are running on.
In contrast, robotic software is active, and does try to understand its input instead of simply processing it according to arbitrary rules. The output of the software has a direct impact on the performance and survival of the robot in the real world - a software error could cause the robot to be squished by a car or lose power if it fails to find the wall socket.
There is no passivity here - robotic software must "push back" on sensory input, whereas PC software must passively adjust to user input.
This difference explains the differences in perception of robots and PCs. PCs are the ultimate slave - trying to do exactly what the user wants with no "personal" input. This makes PCs very flexible but utterly dumb - since they don't try to understand the task they are performing, they screw up left and right, and must be brought back by the user. Even with a novice at the keyboard, the PC is the idiot in human-computer interaction, mostly because it has no idea of what is going on.
In contrast, a robot is not quite a slave, despite the popular image of them as mechanical servants. To qualify as a mobile robot, the device must do a lot of processing that does not consist of orders from a human - primarily sensory data roughly equivalent to human senses. The order from a human is an occasional thing in the overall processing.
One might give an order for a robot to turn left. This will be a small amount of the total processing, which will be processing environmental data so that the turn can be executed. The order is not arbitrary to the robot - both human user and robot attempt to understand the real-world environment in a way allowing a turn left.
The profound difference between PCs and robots makes the typical tech guru blind. Despite the fact that their desk hardware could make a credible robot a la PC robots like White Box Robotics, they know little more than a non-tech person about the industry. The articles don't contain tech details, but instead contain commentary by robotics professionals on the future of their field. Someone who puts in fence posts for a living could collect the same information with the same competence. The PC tech guru - despite all their hardware and software sophistication - is on the outside, looking in.
There's a precedent for this. Imagine it is the year 1978 and you're reading Popular Electronics magazine. An occasional article begins popping up about homebuilt personal computers, and attempts to create useful mass-market PCs by startup businesses. Despite the electronics orientation of the magazine, the tech writer doesn't really understand PCs any more than you do - they are reporting on an areas outside their realm. They don't have a homebrew computer or even an Apple I on their desk. This is despite the fact that the homebrew PCs use the same components found in many of the electronics projects described by the magazine.
It is interesting to watch the future of PC-based publications and gurus unfold with the rise of the robots. Will tech gurus continue to watch the rise of the robots as we might watch a report on Mars exploration? Will new robotics gurus such as those being showcased on Everyday Robots become the new tech luminaries? Will the PC industry, along with its trade publications, celebrities, stock market darlings, tech gods and industry centers, be sidelined as the robotics revolution moves forward? The gap between PC and robotic ideas - despite the common hardware and software - argues that the robotics revolution will happen outside of the personal computer world.
