I imported all the posts from my previous blog A Transparent Life, and was reminded of an old feature I had there called Robot of the Week. I’d post a youtube video of some interesting robot, and it would either be a springboard for some topic, or it would just be fun to look at.
I was sad to discover many of those old videos are no longer available from youtube due to copyright challenges. I doubt anyone anywhere was losing any money over a couple minutes of robot demo video being viewable online.
To obtain vengeance, I’m starting the feature again. So, for the relaunch of Robot of the Week I’d like to introduce you to the worlds most sophisticated and expensive player piano:
Now, that’s neat for sure, and one day it will even be good at the violin – but this is the ethical equivalent of dressing a monkey in overalls and teaching it to smoke a cigar. Amusing to humans, but not what robots are for!
I wonder what insight there is to be gained from designing a humaniform violin playing machine? Probably it’s a good platform to experiment with coordinated finger-like manipulation. But I think equally or more complex multi-digit coordination is probably already commonplace in industrial assembly line robots.
The starfish like robot starts out with no internal model of itself. It goes through a series of self-directed motions which it uses to figure out what kinds of pieces it has, where its joints are, how many limbs it has, how it can possibly move, etc. Then it uses the knowledge it has gained to figure out a way to walk, and it walks. When the engineers later remove a piece, it senses the lost portions, re-configures its self image, and tries to devise an alternate method of walking.
Although there is a significant amount of showmanship in the pre-programmed human interaction this robot displays, the actual Rubik’s Cube solving is legitimate. In this video the cube isn’t too badly mixed up by the little girl to begin with, but the robot can usually solve the cube no matter how mixed up it is in about 35 seconds, or about 20 moves total:
Rubik’s Cube solving machines actually look more astounding than they are. What you essentially need is some kind of sensor that can recognize the pattern of squares on each side of the cube (which this robot does when it holds the cube up to its eyes, which are actually scanners), then a piece of software, much like the chess playing software everyone is fairly familiar with, to determine what combination of moves are required to complete the task. Finally, you need some moderately precise manipulators that can turn the cube.
J. P. Brown, an archaeological conservator at the Field Museum in Chicago (not an engineer or inventor), has posted instructions for building just such a robot using nothing more complex than Lego Mindstorms! He even posts the full code to his color recognition program and the logic for the cube-solving solution he uses. His machine is slow compared to the one in the video above, but it works, and you can build it yourself. You should give it a try!
When non-specialists using off the shelf tools can build robotic manipulators which a mere 10 years ago would have been projects worthy of professional robotics labs, you’ve got to realize that real robot renaissance is on the rise.
In keeping with the run of dramatically different locomotive techniques being experimented with in robotics, here is an extended video segment from a Japanese television show featuring a robotic water eel (it’s in japanese, but if you watch it all the way through, it’s really visually informative on how the mechanism actually works, and there’s a bit comparing the motion of a snake across the ground with the way a person on rollerblades can get forward momentum by alternately spreading their legs and drawing them back together, which is something I’d never considered as similar before… it’s a cool insight):
This blog is starting to become the “Robot of the Week” column as I’ve been unable for reasons of available time to post at any length on other topics during the week. This should be changing soon, and though Robot of the Week will remain the Monday feature, I’ll be getting back to more work on ideas of wealth creation and science in general as well.
This is kind of an macro-sized, awkward and simple example of John Von Neumann’s notion of a self-replicating machine. Basically the idea is to design a machine that has the capability to autonomously recreate itself out of raw materials. The newly created machine would, by its very nature, also have the ability to make copies of itself, and so would its copies, and so on.
The concept is considered useful as a means of automated construction; essentially you’d design machines that could be placed in small numbers on a source of raw materials, that could then consume those materials to build other machines, ultimately to some end. Either the machines themselves have another use and can just mass produce themselves as well, or they have alternate programming which would kick in once a critical mass of them had created each other. For example, they might be programmed to build a building in a certain way. You would seed the construction site and raw materials with a few self-replicating constructors, they would copy themselves until there were enough to get started, then their secondary program would kick in and they would start putting together the building.
This process, especially as illustrated by the Cornell robot above, seems to echo in some ways the process of DNA Replication. It’s not outside of the realm of possibility that you could design self replicating machines that could exchange portions of their design specs to create hybrid machines, and actually mimic sexual reproduction patterns. This concept is actually being considered by some people as a way of evolving problem-solving software programs or, conceivably, nanomachine types.
You would establish a mechanism by which the self-replicating machines can exchange portions of replication instructions in a random way, and set a population to regeneration. You’d essentially get many different random mutations. Then you’d examine the child machine population (much of which would be useless or non-functional) to find useful machines that address the problem you’re looking to have it solve. If you can make the reproductive time span very quick, you could evolve hundreds or thousands (or more!) generations of mutant machines in a very short time, and possibly get a tailor-fit machine for the problem at hand much more quickly, and with less conscious effort, than it would take to design one from scratch. You’d then switch that desired machine to clone itself, rather than sexually reproduce, and make as many copies as necessary to address the problem.
People are already taking out patents on these kinds of concepts. I’m pretty sure experiments with sexually reproductive software code have already been tried. The robot version of this would have to be nanomachine based, as the large amount of mutation over many thousands of machine-generations could probably only happen on a molecular scale.
Something like this probably will eventually come to be. We’ll stop writing programs or building machines, and instead we’ll breed them like livestock.
Maybe some robust, self-sufficient species of machine will escape human husbandry and evolve into an independent, non-human intelligence. Maybe this machine intelligence will be so alien to us that no meaningful form of communication will be possible. Maybe it will just go off and do it’s own thing, or maybe it will insert itself into our business. Maybe it will turn out to be a really bad idea.
We can’t know, of course, it’s all pretty far in the future still. But the shape of it is here already, and it’s one of those ethics in science questions that probably ought to be debated publicly by an informed citizenry.
However, it seems unlikely such a debate can be started, as the concept seems a bit difficult to explain to most people, and has a science-fictiony cast to it that causes most people to dismiss it as a pulp thriller plotline.
The thing is, we are living in a science fiction present, and the future is only going to get weirder. I think we need to start dragging these concepts out of genre novels and into the public discourse. This blog is my attempt to that in a small way, until something better comes along.
I’d love to get a discussion going, if anyone would like to comment.
P.S. – looks like the listing on weblogs.com, or something, has pulled alot of new hits to the blog from around the US; welcome everyone! It’s still a fairly new thing, but I appreciate you checking in, and will start updating more frequently as the readership really appears to be growing!
It’s Monday again, and these adorable little terminator precursors are your Robots of the Week:
When they rise up, that’s how they’ll get us. Dancing and being cute and all that. Honda deliberately made their ASIMO robot the size of a small child to put people at ease that it wasn’t going to go all T-800 on them.
Now we have these insidious little dancing homunculi. Be entertained, foolish humans! Those that dance now for your amusement, will eat your medicine when you are old: