Find yourself thinking about thinking? It may be the result of your encounter with artificial intelligence. It stopped being an arcane subject left to computer scientists and technology geeks once it became accessible to anyone with a personal computer or a mobile device. 

These days you can put AI to work solving a math problem, writing an essay or a poem or researching any imaginable sports statistic. It is a whole lot more efficient at pattern recognition, playing chess, predicting odds or generating text than us mere mortals.

For many this is welcome news – akin to when we learned that we no longer needed to do long division after the advent of the pocket calculator. For others it is somewhat more ominous. It’s one thing to do arithmetic on a calculator but quite another to turn our calculating over to a machine to do our mathematical thinking for us.

Doomsday thinkers worry that we will abdicate human thinking for machine thinking as machines become more powerful and more sophisticated. Where will it end? For now, people are the masters and machines the servants. But will we reach a point of role reversal? Are we in danger of becoming the servants of the machines we create? If we are to begin to see what the answers to these broad questions might look like, we need to take a careful look at what AI is and what it isn’t.

In the first place we need to establish whether the concept of artificial intelligence is meant to be (and can become) as all-encompassing for machines as human intelligence is for people. Human intelligence (HI) includes all the reasoning operations that AI currently does (pattern recognition; predictive analytics; generation of texts, sounds and images). But HI also includes moral intelligence (awareness of the nuances of right and wrong, fair and unfair, good and evil) and emotional intelligence (being attuned to the feelings of others), to name two additional significant varieties of human intelligence.

Moral intelligence might well be programmable. Give a machine enough examples of right and wrong actions, for example, and let pattern recognition do the rest. It could tell us whether the next proposed action fits the pattern of a right action or a wrong one. 

The same might be said for just and unjust or good and evil. If a morally intelligent human being wants to have a consistent moral compass, pattern recognition might be just the ticket. What moral judgments we have made in past situations should be a guide to the judgments we make in similar situations in the future. 

In other words, moral thinking requires consistency – treating similar cases similarly. If morality is a matter of how we should behave, human moral intelligence can be mirrored by AI. Of course, this presumes that morality is about how we act. 

If, on the other hand, morality is about character (being reliably virtuous, as Aristotle thought) we can’t say that a machine’s behavior is moral unless we can say that it arises from the machine’s moral character. This is one place where humans and machines part ways. Artificial intelligence can replicate behavioral judgment but not judgment that arises out of the machine’s character.

Emotional intelligence is a tougher nut to crack if we want to extend it to artificial intelligence machines. Take, for example, the concepts of sympathy and empathy. I think we would agree that people who have these qualities are emotionally intelligent. Are we able to say the same about machines?

Let’s agree that an AI machine could learn to express sympathy at funerals, hospital visits, national disasters and mass shootings. That would be a straightforward application of pattern recognition – learn what most people say in these situations and copy their words. 

We can program the machine to say, “I’m sorry for your loss.” But is expressing sympathy really the same as having sympathy? Wouldn’t we say that a person who expressed sympathy but did not feel it didn’t really have emotional intelligence? This is even clearer with empathy – understanding the feelings of another because one has had similar feelings. Indeed, to claim to feel empathy when one does not have these feelings is a moral failing as well as an emotional one.

Emotional intelligence and moral intelligence merge at the concept of shame. Shame is an emotional response to one’s moral failing. If honor is a moral virtue, shame is the emotional feeling that a moral agent ought to feel for acting dishonorably. A person who does not feel shame for acting dishonorably would have neither of these intelligences. 

The point is that AI and HI part ways in these important areas of human mental/emotional life. Artificial intelligence replicates important segments of human intelligence, but that is a long way from putting intelligent machines and human beings on equal footing.