Saturday, July 22, 2017

Learning The Way That Machines Do

            This article focuses on a series of areas in modern technology that are very complex, very sophisticated and very nuanced particularly for a non-technology person like myself.  It is in dealing with an area like this that I sometimes have wondered what a non-technology person like myself is doing writing a column on the effects of modern technology on living environments and human behavior.  But after having given some thought to this question, I have realized that the key word for our purposes here is effects.  I am not so interested in the operations of modern technology, but rather in their impact on human experience and their capacity to transform human behavior and the very nature of the human sense of self.  I don’t have to be an expert on modern technology in order to perceive and understand its effects on me or on others when we are around it.

            And certainly one area of modern technology where its impact is going to be increasingly felt on humans is artificial intelligence.  This is an area of modern technology which literally tries to recreate human intelligence and even improve on it with machines.  Initially, human intelligence techniques were created directly by programmers through a process called hand coding.  In other words, a human or humans were involved all through the development of a process that imitated human intelligence.  Then along came a new subset of artificial intelligence called machine learning where a human, like the grand clockmaker in the deist concept of God, creates an algorithm and generates the data that then allows the machine itself to create the process that imitates human intelligence.  Finally, there is a subset of machine learning called deep learning wherein layers of artificial neural networks are created that imitate the neocortex of the brain and that imitate human intelligence on a higher more complicated level than was possible with previous machine learning.

            There are a lot of people who are really obsessed with playing God and creating their own complex behavior entities just like God supposedly created them, the human creators, as humans.  But will people today truly create machines and robots that are mentally like people? This is an important question because the identity of humans as unique complex behavioral entities is being chipped away at by all the scientists and engineers that are working constantly to improve the performance of machines and robots.  My thought is that as long as these people continue work on stimulating these entities with the defined discrete stimuli of digital signaling, they will never recreate humans.  And this is because science has made the assumption that all stimuli can be successfully converted into controllable measurable defined discrete stimuli for purposes of scientific investigation, the manipulation of humans and for engineering inventions.  As was discussed in many previous articles, there are lots of stimuli in nature with blurry boundaries that flow across parts of our fields of experience.  We don’t experience our primary contacts with the world as filled with the pixilation of defined discrete points.  If anything, when we are relaxed and not focusing, we experience much of our visual field as filled with entities and patches of substance that flow together.  In other words, individual flowing blendable continual stimuli can agglomerate and form larger stimuli.

            It was also discussed previously that, although there is an infinity of defined discrete stimuli, there is a larger infinity of flowing blendable continual stimuli, stimuli that can be vaguely described, but not in a way that easily distinguishes them from one another.  The infinity of defined discrete stimuli is what one would call a kind of delimited infinity, and the infinity of flowing blendable continual stimuli is what would be called a non-delimited infinity.  I am using delimited and non-delimited in a way analogous to the way they are used in math.  In math, an example of the delimited entities discussed in infinity theory is all the discrete numbers and an example of all the non-delimited entities is all the points on a line let’s say going from 0 to 1.  It can be shown through a mathematical proof that there are a greater infinity of points on a line than discrete numbers.  I make the analogy between the mathematical proof on the one hand, and the comparison between defined discrete stimuli and flowing blendable continual stimuli on the other, because both deal both with entities with defined boundaries and entities with blurry boundaries.  It just makes intuitive sense to me that what is true with different kinds of mathematical infinities can shed some light on the spatial categories of human sensory experience.

            In order for humans to create, control and manipulate their own complex behavioral entities, they have to reconfigure their field of experience to highlight data-defined discrete stimuli, which are more controllable because they represent reduced infinities of possible stimuli.  However, at the same time, humans are very limited in their capacity to instill into these entities more nebulous states of mind: creativity, emotions, religiosity, a coherent sense of self.  None of these things can be implanted in complex behavioral entities on the basis of defined discrete stimuli alone.

            Anyway, digital technology, the foundation of artificial intelligence, machine learning and deep learning is built on defined discrete stimuli.  As the digital technology involved in these machine activities becomes more complex, the behavior manifested in these machine activities becomes more and more complex.  Particularly deep learning, which is meant to operate to some extent like the neocortex in the brain, seems to imitate human cerebral activity.  But however complex it becomes, the activity is still going to consist of a series of defined discrete steps similar to the way a machine operates.  It will not be derived from the flowing blendable continual stimuli that are the foundation of a coherent human sense of self.  The activity will never be based on an irreducible organic agency, will never be guided by an irreducible cerebral consciousness, will never be guided towards making, receiving and preserving organic imprints, will never be accompanied by an irreducible experiencing of what is happening and what the machine is participating in, and will never have a coherent reflexive awareness of mortality and a desire to prepare for death through a surrogate immortality based on preserved organic imprints.

            But as particularly deep learning increasingly approximates human activity, people are going to increasingly blur machines and humans in their minds which is going to lead to a diminishing of the appreciation of the uniqueness of humans.  But humans should never be reduced to machines, because, however skilled machines become, even in some cases improving on human skill, there are certain ways that humans will always be so much more than machines, and that is something of which we should never lose sight.

(c) 2017 Laurence Mesirow                      

No comments:

Post a Comment