Overview

Clark’s work serves several goals. The first of which is a review of contemporary Artificial Intelligence as applied to robotics and neuroscience. Clark’s second goal is a critique of traditional symbolic AI and the notion of disembodied reasoning that pervades it. The brain, body, and world all tie together in a densely connected network that is impossible to untangle. Thinking and learning are strongly rooted in the mechanics of the body and world. These systems are adaptive and pattern based, not symbolic. Clark specifically criticizes projects such as Cyc, which attempt to build a filing-cabinet model of the world by extensively cataloging each relation and bit of information.

One of Clark’s first examples of embodied reasoning is the cockroach, which is an efficient and well adapted creature, and has a bare minimum of computational reasoning in its tiny brain. This computation is directly tuned to the roach’s physical environment.

Notes

Clark critiques the use of symbolic reasoning in towards the application of embodied problems. In relation to a simulated world, where everything is symbolic, the situation gets rather hairy. Clark’s preferred approach may be to avoid simulation entirely, but that is not an option here. Nonetheless, he has several points of merit that are applicable in the simulation of systems.

Clark repeatedly emphasizes that creature behavior evolves in relation to the physical environment as well as its physical body. Brains are messy systems that leverage as much as they can off surroundings and affordances. They are not planned, but are highly adaptive and responsive systems.

With this in consideration, it should stand to reason that the most successful AI simulations of people are ones in which the characters leverage their environment as much as possible. The Sims leaps immediately to mind, since most real logic is represented in the objects in the environment as opposed to in the Sims themselves.