Automatic Design and Manufacture of robotic life forms
Lipson and Pollack, Nature (2000) vol 406:974-978
Robots ever more brainy; K.Cheng, San
Diego Union Tribune, 8/31/2000
The robot that made itself – almost....
The article ‘One giant leap for machinekind?’ (Los Angeles Times; 8/31/00) tells the story of self-evolving and self-generating machines comparing the process to biological evolution. The original report published in the British journal ‘Nature’ presents an astonishing story of ingenious engineering, programming a computer to evolve a machine designed to move around (see Golem project at Brandeis University). While not interfering in the manufacturing process, the authors Pollack and Lipson undoubtedly put in the best in information technology they have to offer. As designers, Pollack and Lipson admit that theirs is not a self-replicating robot, nevertheless compare their design with that of living organisms.
Here is where my objections start. Behind all the excitement is a misrepresentation of facts of not distinguishing between design and biological evolution. There is a difference between a computer instructing a machine (with a great name, by the way) to 'give birth' to a virtual robot on the one hand and a living organism on the other hand that replicates through a single cell growing into an adult form producing yet another single cell growing into an adult form - true self-replication found in all organisms from animal to plant to bacteria. The use of the word ‘design’ in the original article (Pollack and Lipson use it twenty times), which by the way is almost never used in the newspaper story, is testimony of this lack of self replication, of blind (non directional) evolution instead of ‘forced evolution’. Pollack gives instructions to use three type of building blocks and to select for movement. Movement, however, has nothing to do with the (non existing) self-replicating ability of the robot, but is a given ‘fitness’ parameter by the designer. Function and reproductive mechanism must be coupled. The different moving robots do not really compete with each other, but are selected by the designer, who never happens to change the selection criteria. Darwinian evolution is based on the ability to reproduce, not movement per se. But since self-replication is not part of the design, it cannot be compared with biological evolution.
To a biologist this heavy handed reference to ‘design’ while constantly invoking evolution and replication comes as a shock. The use of language here is better suited to support a creationist account of evolution than one of natural selection. Clearly, there are superior beings behind the design of the robots, so clever, indeed, that these beings managed to write a software program to simulate evolution of a movement. These superior beings then buy an off the shelf machine (built by other superior beings) which manufactures a robot replica in the real world (a process claimed as independent of human intervention). From the entire design and manufacturing process, there is nothing reminiscent of a biological process. The evolution is virtual, the machine that produces the robots, has not evolved itself, and does not create anything resembling itself. This is an example of a master designer, not a self replicating unit. Long live creationism.
Computers are very good tools to simulate reality and the availability
of super computers has drastically improved life-like appearances of many
things virtual (e.g. dinosaurs in movies). To confuse this with biological
processes reflects a lack of comprehension of biological complexity or
the nature of metabolic processes. How else should I understand the quote
from Pollack equating these robots (‘primitive nature of their creation’)
to the supposedly low complexity of bacteria. Either he has been quoted
incorrectly, has no idea of the complexity of these single cell organisms,
or the word ‘bacteria’ stands as a mere metaphor to explain to the non-scientist
(journalist) the concept of a primitive organism.
Return to IN THE NEWS ....