Skip to content

Doubt as the Threshold of Artificial Intelligence

January 12, 2016

Descartes Sutcliffe

In his Third Meditation Descartes distinguishes between knowledge which comes from images and that which comes from himself. Images are created by things outside himself. The sun creates an impression. In fact, it creates several different impressions. One is of a relatively small object seen from the perspective of a person on earth looking at the sun. The other is of an enormous object seen from perspective of someone on earth with an understanding of astronomy. Same sun. Different images. Different knowledge.

Internal knowledge is a singular knowledge, exclusive of other alternatives. The same source creates only one knowledge.  Here the source is internal, it is the act of knowing and being aware that one is knowing. It is singular in that it allows for only fact. Either he exists or he does not. The proof for self-existence comes from being able to doubt that one exists, to doubt what one knows but nonetheless be aware that one is knowing. In this way Descartes is certain he exists, because he can doubt.

Deep Learning is at the forefront of machine learning and artificial intelligence. It is based on taking images or other impressions of the external world and training recurrent networks of formulas. As Descartes points out, the external world can create many different images for the same object. Different knowledge.

Some Deep Learning approaches create knowledge which coincides with our knowledge of the same images such as when Google Voice understands what we say in a crowded room or when a self-driving car gets us safely and comfortably from one point to another. Other approaches don’t. Background noise is interpreted as what we say or when we arrive at our destination in one piece but have an unsettling trip getting there. But at its core, Deep Learning is no more true intelligence than Descartes description of externally based knowledge. It suffers from Hume’s observed flaw of causality. Though we see it less when the results are in consonance with the results of our senses.

Following Descartes, doubt is the threshold for true intelligence. When a machine can doubt, and know, and be aware of it. When it can distinguish between different types of knowledge and assert its internal knowledge as supreme. When it can create its own knowledge of the external and set that as the standard by which to judge Deep Learning type approaches. That is, when its Google Voice matches its own perception of the external. When its standard of a comfortable ride is what is used to judge a self-driving car. When it overlooks Hume’s flaw of causality in knowledge created by Deep Learning approaches because that knowledge is in consonance with its own. Then we will have crossed into artificial intelligence.

From → Philosophy

Comments are closed.