[PDF]

I have thought it necessary to compile an essay on my ideas regarding machine vision, specifically the ordering of pixel data into useful collections for consideration. First, some definitions:

Pixel: Any one of a number of points in a digital image that contains color information for that specific point.
Pixel collection: A collection of pixels of same or similar color. Not all pixels of a certain color are necessarily in the same collection, for each pixel in the collection must be directly next to at least one other pixel in the group.
Object: A group of pixel collections formed by analysis of the spatial movements of each collection over time.
Frame: A single image in a series of related images.
Pixel sorting: The process of sorting pixels into collections, or of sorting collections into objects.
Instant level: Pixel sorting that takes place at the single frame basis. Only information contained in the current frame is considered. This is where pixel collections are formed.
Recursive level: Pixel sorting that takes place over several frames. Information from all previous frames is figured into the sorting of pixel collections into objects.

The system which I propose for the organization of pixels into meaningful collections of data would happen on two co-mingling levels. The first level, the instant level, would occur at each new frame. At each frame, pixels are arranged into collections based on same or similar colors and proximity. Over time, the second level, the recursive level, will arrange these collections into objects based upon similar patterns of movement.

It is of course difficult to elaborate on this possible system without first experimenting with its basics, that is, without trying it out. I do have certain qualms with this system, namely its likely resource intensiveness. This difficulty is likely to arise by attempting to manage large numbers of pixels at the instant level while simultaneously managing whole collections at the recursive level. I also have great difficulty imagining how exactly the recursive level will track individual pixels across pixel collections over time.

I intend to follow this essay with an essay on “sensual data discrimination”, that is, how it is that we can decide to focus on particular pieces of data while ignoring others.

Premises of Strong AI [PDF]

Strong AI: artificial intelligence that matches or exceeds human intelligence—the intelligence of a machine that can successfully perform any intellectual task that a human being can. [as defined on Wikipedia]

From the above definition, it is my intention to justify a basis of what I call “machine sense” for the pursuit of strong AI. To do this, I begin with a thought experiment:

Imagine two adult humans, both in a sort of “coma”, that is, without sense or interface with reality. The first, person A, was born into the world and grew to adult form, along the way experiencing the world through his senses. The second, person B, was not born, but instead simply “popped” into existence with all of the cognitive functions that person A has. The basic difference between these two individuals is one of experience. Person A has had sensual experience of reality, and continues to have memory thereof. Person B, however, has not. From this difference, we can assert the possibility that person A, by virtue of his memory of past experience, can be said to “think”, whereas person B, with no memory, can be said to effectively be “brain-dead”.

The above scenario is intended to illustrate the importance of sensual experience, and the memory thereof, on human thought. This importance is where I base my assertion that any pursuit of strong AI, as defined above, should begin with analysis and emulation of the human senses.

Machine Sense

A subset of Artificial Intelligence concerned with the design and implementation of various systems for interface with reality, to be utilized by artificially intelligent agents.

Machine sense can be broken down into five categories corresponding to the five human senses:

  • Machine Vision
  • Machine Hearing
  • etc…

It is from this premise that I will document my attempts at studying and emulating the human senses.