Sunday, March 18, 2018

Devising Moral Behavior For Machines

            Every time this column comes along to try and draw the boundaries between the  respective essences of humans and machines, computer engineers create new situations where computers seemingly adopt another trait of humans that further blurs the distinctions between these two kinds of entities.  In a recent article, there was a discussion about how artificial intelligence (AI) has developed to the point where computers can beat humans in games like chess, Go, and video games.  The common denominator that these games have in terms of appropriate behavior is that they all require adversarial competitive behavior.  In an article by Rich Hardy in the online publication New Atlas on Jan. 21: “Morality algorithm lets machines cooperate and compromise better than humans”, there is a discussion about some computers of some computer engineers who have created a morality algorithm which allows computers to cooperate and compromise with humans in the performance of certain tasks.  So now humans can kumbaya with computers.  At least that is the ostensive interpretation of the title of the New Atlas article.  But the question is does cooperation and compromise mean the same thing to a computer with artificial intelligence (AI) that it does to a human?  What does it even mean to talk about cooperation and compromise within the context of a computer with AI?

            First of all, a computer is a machine, and it is defined by all the defined discrete inputs that it receives and all the defined discrete processes that it carries out.  Algorithms set up an instrumental basis for operation of some very advanced machines.  It can be a very complicated basis for operation and it can demonstrate some very complex activity that people who want to anthropomorphize particularly some computers would call complex behavior.  But algorithms are process-oriented, and they don’t provide a basis for operations based on some unique essence.  The unique essence underlying behavior in humans is an organic sense of self.  The organic sense of self is based on components that deep-bond with one another to form a larger indivisible organic whole.  These components are held together functionally by the flowing blendable continual stimuli that have been discussed in many different contexts throughout this column.

            In other words, all the algorithms in the world, however complex they may be, are not going to be enough to create an organic sense of self in the sense that humans have.  Without this organic sense of self, there is no real sense in which computers are really cooperating or compromising with humans.  To use the terms cooperating or compromising with computers represents a form of anthropomorphizing that distorts the interactions that are actually taking place.  There can’t be any real cooperating or compromising taking place between two entities, unless both parties to the cooperating and compromising are coherent wholes, activated not only by the defined discrete stimuli measured in laboratory experiments, but also by the flowing blendable continual stimuli found in life.  From this point of view, a morality algorithm is a meaningless concept.

            But what are the implications of thinking in terms of a morality algorithm?  Rather than elevate the behavior of the artificial intelligence (AI) on the computer, it may diminish the importance and dignity of morality behavior among humans.  The behavior of humans and AI in computers blur together in human minds.  Our moral actions are increasingly seen as simply proper discrete instrumental processes generated to create a sense of stasis or balance in the interactions we have with the complex behavioral entities that we call computers.  There is less emphasis in our thinking on the core centers in our minds from which our moral actions are generated.  And yet without these core centers in our minds – our organic senses of self – there is no meaningful entity to which the label moral can refer.

            The only conclusion that we can draw from the continued false parallel between humans and computers among our modern computer engineers is that a meaningful core center in an interacting complex behavioral entity is no longer important to them.  What is important to these computer engineers in defining life or pseudo-life is simply instrumental process and not an essential organic core center.

            And, of course, these attitudes are spreading to other sectors of modern technological society.  More and more companies avoid paying benefits and pensions to their workers.  Workers are increasingly no longer thought of as coherent wholes with organic senses of self, with organic essences, entities that derive their value not just from what they contribute in the processes of their work, but from their intrinsic worth as human beings.  Instead, they are looked at in terms of their relatively short-term instrumental worth, performing certain labor until their immediate projects are completed, until they are no longer needed for a specific kind of work in a company, until they are no longer physically or mentally able to do the work or until they reach an age cutoff within the company after which people retire.  But in many companies, people are no longer considered to be human beings that are to be treated with dignity when they retire.  Rather they are looked at as machines that are falling apart, that are considered to be towards the end of their usefulness, towards the end of their general worth.  The obsolescence of these humans in more and more cases leads to their being disconnected from the company when they retire, like machines that can no longer perform their tasks and are to be discarded.  Actually, because AI can be kept alive seemingly forever, humans are increasingly going to be compared to more primitive machines.  More primitive complex behavioral entities.  And now that AI is being improved with the morality algorithms, an important superficial behavior application is being added which allows computers with AI to be considered almost collegial.  So in the blurring of identities, humans are being reduced to a more primitive less lasting machine status, and certain machines are being elevated to a status that is increasingly in certain ways perceived to be superior to human status.

            What has to be always remembered is that it is the organic core that matters, the organic sense of self.  All the complicated morality algorithms in the world may be able to fine-tune computer interaction with people, but it is all related to instrumental matters and never matters of essence.  What all is said and done, the computer is being instructed that when situation a occurs, do b.  A series of programmable possibilities.  This by itself does not create a human.

(c) 2018 Laurence Mesirow

No comments:

Post a Comment