Consciousness Explained
Daniel C. Dennett, 1991

Artificial Hallucinations
In the
Prelude , I believe Dennett has missed a major point about the simulation of a high quality hallucination. As the point of stimulation varies up the pathway of perceptual processing from the outer periphery to more central points, the "combinatorial explosion" of required stimulation is progressively unwound as each level of perceptual analysis is bypassed. He is quite correct that the task would be essentially impossible at the level of a virtual reality machine (of today's technology) but dreams and hallucinations most certainly are not "injected" into the perceptual stream at such outer peripheral levels.

If, instead, the stimulation is applied at a higher level, where the detailed sensations have already been decoded into a single representation in the form of a symbol for that sensation, the task becomes much easier. The symbol pattern for the sensation could be activated, in some cases by a single binary level, in other cases probably requiring a graded stimulation.

This is not to claim that a final level exists at which point a single neuron could be stimulated to reproduce the effect of sand flowing through the fingers. It seems more likely that it would be something like a half dozen or so effects, a light pressure at particular skin areas and perhaps a coolness, fluctuating at rates certainly no greater than 10 or 20 changes per second -- easily within the power of a small embedded microprocessor if applied at the point where the corresponding sensation has already been decoded.

A Virtual Machine?
A virtual machine architecture may be as simple as what
Newell and Ballard describe as a heirarchical system.


Top of Page | CE Summary | Sort by Topic | Sort by Title | Sort by Author