Templeton Research Lectures - The Center for the Study of Religion and Conflict

Below we will feature a frequently updated collection of essays from scholars in the sciences and humanities on human identity and enhancement. These will be reflections on human identity relative to technological enhancement and the potential impact of human enhancement on our identity and the social structures built around the various understandings of what it means to be human. The essays will, for the most part, be solicited efforts but we are willing to consider unsolicited submissions. Please feel free to contact us here.

Programming Compassion: A Challenge Ahead in Strong AI by Sean A. Hays, Arizona State University

In the March 20, 2007 edition of the New York Times there appeared an article claiming that that the precursors to human morality had been discovered in chimpanzees. The third paragraph of this article describes an experiment where a pair of monkeys were placed in an experimental device that required one monkey to pull a chain to receive food, but when the chain was pulled it would also deliver an electric shock to the monkey's companion. The result was that the initiator of the shocks eventually starved himself for several days rather than electrocute his companion. The article relates this experiment rather uncritically, I fault the reporter here for failing to offer even a brief mention of the idiocy and inhumanity of such an experiment, and moves from there to other examples of non-human primate compassion. Compassion, and empathy, are supposed to be precursors to moral systems. I find the overall claims of the research to be rather dubious in so far as they point to an evolved morality, a prospect I find much less likely than our having underestimated the sophistication of the social systems we observe in other animals.

The more significant point for our purposes is the briefly mentioned experiment with the monkeys. What that experiment demonstrated was that the monkeys did seem to possess some degree of compassion and empathy, the experimenter clearly DID NOT. Now, there are two possible explanations of this "scientist's" complete lack of humanity and decency, not to mention the incredibly poor experimental design and the subsequent lack of any meaning or strength in data so obtained, and they are sociopathology or the socially conditioned human predisposition to only feel compassion for other humans, or significantly less compassion for other non-human animals. If the experimenter is a sociopath, a very real possibility given the nature of the experiment, then we would expect that he would have felt no greater degree of empathy if it had been his neighbor's children in the device, shocking and starving themselves. If, however, this experimenter simply possessed the conveniently scalable empathy that is common among humans which allows us to justify the brutal torture of other animals by claiming that they are less valuable because of their inferior intellect or lack of human emotion and cognition would it make him any more acceptable as a neighbor? I'm being intentionally provocative here but I would like to make clear that I deplore this anonymous researcher in reality despite his utility in demonstrating the point I am about to make.

What does all of this have to do with transhumanist technologies you might ask? It relates to a problem I have noticed in the descriptions of strong AI (artificial intelligence) offered in Ray Kurzweil's The Singularity is Near: When Humans Transcend Biology. Kurtzweil asserts, most often in the amusing fictional dialogues he includes at the end of each chapter, that machine intelligence in the future will revere their biological predecessors. It is never adequately addressed in the book why they would choose to revere us and I find this to be terribly problematic given Kurzweil's familiarity with the field. It indicates to me that AI researchers are not spending considerable time thinking about how we might program the machine intelligences of the future to have compassion for their idiotic ancestors.

I am not, typically, a reactionary and I am not, I hope, being one now. I think strong AI is an inevitability and I find Kurtzweil's incorporation of exponential growth curves and the law of accelerating returns to be convincing (though I think he often underestimates the speed with which future technological innovations are approaching). I think it is possible to create artificial intelligence in a way that will result in it being principally a tremendous social good and, perhaps, an exceptional addition to the human family. I do not, however, have any faith that this will happen without tremendous effort on our part. I prefaced the following remarks with the preceding mea culpas to avoid the charge that I was simply reasserting the common claim that we will eventually become the pets of our machines. I don't think that is likely, if we take the right steps. I think it is fair to make a general analogy between the experimenter and monkeys described above and artificial and biological intelligences of the future. They will massively exceed the capabilities of humans and their superiority could just as easily lead them to adopt the perspective humans have long held toward their less capable cousins. In other words, why should we assume that the massively superior intellects of our future machines will be benign towards us in a way we have never been towards the inferior intellects of non-human animals?

I think it is clear at this point that I believe the assumption that superior intelligence equates to greater compassion and empathy is vacuous at best. Compassion and empathy for those who are less capable than ourselves is a quality that must be taught, and the teaching is often a difficult and thankless labor. Young children do not innately understand the pain they inflict on insects and small animals in their care and must be conditioned to recognize this pain and respond to it appropriately. Our machines will require the same instruction, we will need to program them to be compassionate. We will need to embed deeply in their code a desire to seek greater compassion for the world around them. We will need to program them not only to know how to appreciate Wagner or Dostoevsky but why they should appreciate the art and culture of their less capable ancestors, just as children must be taught to respect other animals through an appreciation for the qualities they possess. If we fail in this regard and push forward with the creation of a generic strong AI that is not designed to value its creator we will become the monkeys who were so sorely used by the scientist above.