Subscribe to feed
Creative, Self-Learning, Multi-Sensory, Adaptive Robots
Wired News has an article (Experimental AI Powers Robot Army) which has a lot of very interesting things to say about self-learning robots and swarm technology.
This is software for generating new ideas on the basis of existing ones, and it has already written music, designed soft drinks, and discovered novel minerals that may rival diamonds in hardness.
The Creativity Machine is Stephen Thaler’s brainchild and has the special feature of incorporating “noise” in to the neural network to help come up with new ideas based on existing ones, and then comparing them to a knowledge base where it can toss out bad ideas.
One example of this, is that a robot was given 2 (or 3) sets of limbs and was able to master self-learned locomotion… that is to say the robot was able to develop the best, most efficient use of the limbs to produce the fastest speed of motion, without ANY programming on how it could, or should move the limbs. In fact, it even learned to run upright on two legs which it found to be it’s fastest mode of travel.
The robots also learned the most efficient way to map their environment, group together in defensive formations, and even sacrifice one of their kind to distract a human in order to meet their goals…. yes, of course it’s part of a military project!
Thaler’s current project, which should be completed over the next six months, will develop a piece of software called CSMARRT (for Creative, Self-Learning, Multi-Sensory, Adaptive, Reconfigurable, Robotics Toolbox). The software can be used to design and model virtual robots that can be placed in virtual environments to learn and develop. This toolbox can create software to control any robotic hardware, handling locomotion, sensors and intelligent behavior to carry out a mission including swarming. The AFRL [Air Force Research Lab] will make this available to its customers — other branches of the U.S. military.
One of the most interesting things about this particular project, is that it involves virtual robots in a virtual environment, which means you can easily simulate hundreds or thousands of robots all working together to solve various problems at speeds of current computing power, rather than the bulky, slow(er) limits of our current robotic technology. Which, in turn, means the robots could be given a task, and then work it out at incredible speeds, trying and retrying all sorts of various combinations until the come up with “the best, most efficient solution”, and then show that solution to people who would actually take that action.
The robots that the Air Force Research Lab is currently working with are the size of cockroaches, and validation tests are said to be classified, but the Pentagon has some new projects starting next year (2007) which includes such phrases as “technology necessary for robotic systems to attack tunnel complexes” and “microdamage technologies” for “very small robotic weapons.”
One of the last phrases to catch my attention was “CSMARRT could potentially turn any robot into part of a versatile swarm intelligence capable of learning and carrying out complex physical tasks.” which made me immediately think of Skynet (yes, from the Terminator movies). What are some of the possibilities if we begin to have artificial intelligence systems that start to self-load this type of software on to other machines, and is then capable of reaching “some self-acquired goal” ? If this software is already capable of having part of it’s mechanical “being” self-sacrifice itself for the survival of the whole, when do we start to say “it’s alive” ? If the goal was to learn how to infiltrate “enemy” computer systems world-wide and take control for some military purpose, giving itself away, or even more scary, to NOT give itself away, and then download it’s software and control every programmable machine it has access to either on the internet or wirelessly…. what then?
No Comments
No comments yet.
Comments RSS TrackBack Identifier URI
Leave a comment
You must be logged in to post a comment.