This is an emerging field that attempts to simulate the behavior of living things in the realm of computers and robotics. The field overlaps artificial intelligence (AI) since intelligent behavior is an aspect of living things. The design of a self-reproducing mechanism by John von Neumann in the mid-1960s was the first model of artificial life (see von Neumann, John). The field was expanded by the devel-opment of cellular automata as typified in John Conway’s Game of Life in the 1970s, which demonstrated how simple components interacting according to a few specific rules could generate complex emergent patterns. A program by Craig Reynolds uses this principle to model the flocking behavior of simulated birds, called “boids” (see cellular automata).
The development of genetic algorithms by John Holland added selection and evolution to the act of reproduction. This approach typically involves the setting up of numerous small programs with slightly varying code, and having them attempt a task such as sorting data or recognizing patterns. Those programs that prove most “fit” at accomplishing the task are allowed to survive and reproduce. In the act of reproduction, biological mechanisms such as genetic muta-tion and crossover are allowed to intervene (see genetic algorithms). A rather similar approach is found in the neural network, where those nodes that succeed better at the task are given greater “weight” in creating a composite solution to the problem (see neural network).
A more challenging but interesting approach to AL is to create actual robotic “organisms” that navigate in the physi-cal rather than the virtual world. Roboticist Hans Moravec of the Stanford AI Laboratory and other researchers have built robots that can deal with unexpected obstacles by improvisation, much as people do, thanks to layers of soft-ware that process perceptions, fit them to a model of the
world, and make plans based on goals. But such robots, built as full-blown designs, share few of the characteristics of artificial life. As with AI, the bottom-up approach offers a different strategy that has been called “fast, cheap, and out of control”—the production of numerous small, simple, insectlike robots that have only simple behaviors, but are potentially capable of interacting in surprising ways. If a meaningful genetic and reproductive mechanism can be included in such robots, the result would be much closer to true artificial life (see robotics).
The philosophical implications arising from the pos-sible development of true artificial life are similar to those involved with “strong AI.” Human beings are used to view-ing themselves as the pinnacle of a hierarchy of intelligence and creativity. However, artificial life with the capability of rapid evolution might quickly outstrip human capabili-ties, perhaps leading to a world like that portrayed by sci-ence fiction writer Gregory Benford, where flesh-and-blood humans become a marginalized remnant population.
The development of genetic algorithms by John Holland added selection and evolution to the act of reproduction. This approach typically involves the setting up of numerous small programs with slightly varying code, and having them attempt a task such as sorting data or recognizing patterns. Those programs that prove most “fit” at accomplishing the task are allowed to survive and reproduce. In the act of reproduction, biological mechanisms such as genetic muta-tion and crossover are allowed to intervene (see genetic algorithms). A rather similar approach is found in the neural network, where those nodes that succeed better at the task are given greater “weight” in creating a composite solution to the problem (see neural network).
A more challenging but interesting approach to AL is to create actual robotic “organisms” that navigate in the physi-cal rather than the virtual world. Roboticist Hans Moravec of the Stanford AI Laboratory and other researchers have built robots that can deal with unexpected obstacles by improvisation, much as people do, thanks to layers of soft-ware that process perceptions, fit them to a model of the
world, and make plans based on goals. But such robots, built as full-blown designs, share few of the characteristics of artificial life. As with AI, the bottom-up approach offers a different strategy that has been called “fast, cheap, and out of control”—the production of numerous small, simple, insectlike robots that have only simple behaviors, but are potentially capable of interacting in surprising ways. If a meaningful genetic and reproductive mechanism can be included in such robots, the result would be much closer to true artificial life (see robotics).
The philosophical implications arising from the pos-sible development of true artificial life are similar to those involved with “strong AI.” Human beings are used to view-ing themselves as the pinnacle of a hierarchy of intelligence and creativity. However, artificial life with the capability of rapid evolution might quickly outstrip human capabili-ties, perhaps leading to a world like that portrayed by sci-ence fiction writer Gregory Benford, where flesh-and-blood humans become a marginalized remnant population.
No comments:
Post a Comment