--- In Wittrs@xxxxxxxxxxxxxxx, "SWM" <SWMirsky@...> wrote: > Is this picture really all that different from Dennett's proposal > that brains run processes in the way computers run algorithms? Here's a quick comparison. I'll use "A:" to prefix the AI/Dennett view, and "N:" to prefix my view. Information A: Information is a naturally occuring part of the world, and is picked up by sensory cells. N: Information is inherently abstract, so does not exist apart from its construction and use by humans (or other cognitive agents. We interact with the world in order to construct information, and we use sensory cells in that interaction. Core functionality A: Computation/logic, applied to the information picked up by A: sensory cells. N: Information gathering, which I shall loosely refer to as "measurement". Starting point A: Most AI people assume large amounts of innate knowledge or structure, perhaps in the form of a program and a data base (often called a "knowledge base"). N: Self measurement of internal states. The system can be said to have, as innate purposes, the maintaining of internal states within innately prescribed limits. Among those innate purposes is a drive to explore ways of interacting with the world, including ways of forming information about the world. Learning A: The usual AI view of learning is one of discovering patterns within the input that is picked up. There is also some consideration of reinforcement learning. N: Learning is acquiring behaviors which tend to promote the ability of the system to meet its purposes. With each new behavior, there is an accompanying new measurement system for self-measuring performance in carrying out that behavior. Of particular importance are behaviors that provide ways of forming information about the external world - we can refer to that as discovery/invention of new ways of measuring. Note that this could be described as perceptual learning. N: With each new way of measuring, there is an associated new concept (that which is measured). With each new self-measurement associated with new acquired behaviors, there is a new purpose of carrying out that new behavior appropriately. Intentionality A: The usual AI view is that there is nothing more to intentionality than attribution. That is, there is only derived intentionality. Dennett argues for that in his "The Intentional Stance." N: The initial self-measurement of internal states, and the consquent initial purposes, are perhaps best considered to be examples only of derived intentionality. However, the new measuring systems created by the system itself are best considered to be examples of orginal intentionality. In particular, information about the world that is formed on the basis of these acquired measuring systems should be considered intentional information. Free will A: The behavior of the system is determined by the input and the mechanistic rules it is following. The system is free to choose only in the compatibist sense that it is free to accede to doing what the mechanism dictates that it shall do. N: Free will is the ability to make pragmatic choices. The options are evaluated according to the systems purposes, and a choice is made in accordance with those purposes. Note that there might be several relevant purposes and some of them might be in conflict. Regards, Neil