Artificial Intelligence

By Dr. Christopher Plumberg

Is artificial good or dangerous? It depends on what exactly one means by "artificial intelligence" (or "AI"). The "artificial" part is pretty unambiguous: it means that we (humans) have created it ourselves, rather than discovered an instance of it in nature. So the really thorny part of this concept is what is meant by "intelligence." There are basically two kinds of "intelligence" that AI researchers distinguish between: weak AI and strong AI. Strong AI is what most people think of when they read science fiction which discusses AI: it means that a computer, created by humans, can do any cognitive tasks (i.e., solve any problems) which human beings can. It involves consciousness and perception of the outside world, in the same sense that God created us to perceive the world. Weak AI, on the other hand, refers to computers which can only perform tasks within a small range encompassed by human abilities. Generally speaking, weak AI machines can only do a few of the things humans can, and then only by way of imitation. Strong AI is only a theoretical concept at the moment; no one even claims to have been able to generate such a machine, and doing so is still a long way off. All modern day examples of AI are of the weak variety, and are not genuine examples of consciousness or self-awareness.

A good example of weak AI is the iPhone application "Siri." Siri is in no sense of the term self-aware or conscious. Rather, "she" is a program intended to imitate the way a human being would traditionally function in the same role. It is truly impressive the extent to which human activity can be emulated by such a program. The essential idea is to mathematically model human techniques of reasoning, and teach the computer to respond in pre-programmed ways to given stimuli. For example, when asking Siri for directions, people will almost always begin the question with "where is..." or "how do I get to...", meaning that Siri can use speech recognition to know when a person is asking a question of this sort. Everything that follows the opening words of the question is then treated as the place that the person need directions to, and Siri conducts a Google search to find out the best way to get to that place. All of this is relatively easy to teach a computer to do, and there are many more techniques for expanding Siri's ability to answer even more difficult questions. The important point to realize is that none of these techniques involves Siri thinking for "herself"; "she" is merely a machine which has been programmed to handle pre-specified forms of human interaction in as general a way as possible. Siri is completely incapable of doing anything except exactly what "she" has been programmed to do, so don't expect "her" to voluntarily order you a pizza any time soon.

So all currently existing examples of (weak) AI are basically just glorified computer programming, and are therefore perfectly compatible with belief in Scripture. Things start to get a little sticky when we ask questions about strong AI. In particular, is it even possible?

Before I try to answer this question from a Christian perspective, let me first note why the idea itself is even plausible to many people. Since most scientists believe that life (including human beings) originated and evolved, unassisted by any higher Intelligence, from completely physical substances by undirected, entirely natural processes, the expectation is that we should be able to replicate this purely physical procedure in some sense today. On this view (known in various circles as naturalism or materialism), the human brain is nothing more than a highly advanced, incredibly complex super-computer which arose ultimately as a result of stochastic processes operating over extremely long periods of time. If this is the case, then one would expect to be able to create other computers like it, given sufficient time and resources. This is the goal of the strong AI program: to create consciousness from purely physical materials and by means of purely physical processes, just as evolution is thought to have "created" human consciousness. So, on the naturalistic worldview, it must be possible to produce consciousness from raw matter, since this is, after all, how we got here.

Now contrast this with the Christian perspective. The Bible reveals that God is inherently communicative: His Son is the Word (John 1:1). He speaks (Deuteronomy 5:27) and creates by speaking (Genesis 1, Romans 4:17). He reveals Himself to us through His Word (Jeremiah 30:2), through His Son (Hebrews 1:1-2), and through His Spirit (Ephesians 3:5). Moreover, Scripture also indicates that we are created in His image (Genesis 1:26), which entails our ability to communicate and respond to the direction of others in free and unconstrained ways. We are consequently distinct from the rocks, plants and animals of God's creation, since we are able to commune with God and know Him (Jeremiah 9:23-24, 1 John 2:13-14). There is therefore an unbridgeable divide between the human and non-human components of God's creation, which cannot even be crossed by human ingenuity: computers can never be taught to communicate or worship, because they are not made in God's image. It follows from this that there exist human, cognitive activities (worship, communication, expression of feeling, sentience, etc.) of which no computer is capable, even in principle. The naturalistic commitment to the goal of strong AI is therefore not founded in reality, and cannot ultimately be successful.

A word of caution is in order: the only way to (empirically) prove whether or not strong AI is possible is to "be" the robot, so to speak. In other words, it's not enough to "see what the computer sees" on a screen and call this "consciousness"; we can only know consciousness on a first-hand basis, meaning that only the computer is capable of "knowing" whether it is conscious (if it indeed were to become conscious). Of course, there would be no way of proving whether this ever happened, since no one except the computer could possibly be the computer or have the computer's experience. The computer could, of course, try to tell us that it had finally become conscious, but there would be nothing to definitively distinguish this from an ordinary machine which had simply been programmed to say the right things and respond in the right way. In short, consciousness can only be observed by the person who is conscious; from anyone else's (external) perspective, it is impossible to tell with complete certainty whether consciousness has actually been attained.

This has motivated some AI researchers to "move the goalposts" a bit: rather than define strong AI in terms of actually achieving perception (which is unobservable in principle), most strong AI research concentrates on varying definitions of "intelligence" in terms of exclusively observable properties. This has given rise to ideas like the Turing test, which defines intelligence as the ability to hold a conversation which is indistinguishable from that of a human being. Notice that the strict goal of "perception or consciousness" has been cleverly swapped out for "something which is indistinguishable from perception or consciousness." However, not only is this kind of "intelligence" somewhat different from what we usually mean by the term, it is even possible for people (i.e., human beings) to fail the Turing test, apparently implying that they are not human! So strong AI research has been woefully unsuccessful at even approximating the incredible creation of the human person that God spoken into being. Indeed, the goal of strong AI is biblically unattainable, and claims to actual progress in this field should be treated with a healthy amount of caution and skepticism. Ultimately, only God is capable of creating a mind, such as is implied by genuine intelligence. So there can be no such thing as true, artificial intelligence, biblically speaking.

It seems likely, therefore, that any claims to strong AI would either be inadequate imitations of the real thing, or would have demonic origins in some way. Since human beings cannot create mind, which is a prerequisite of strong AI, the only remaining possibility would be an unembodied mind (such as an angel) which is capable of deceptively appearing to "inhabit" a mechanically constructed body of some kind (like a robot), implying that the imitation would be demonic. However, I should strongly emphasize that the notion of strong AI is entirely science fiction, at least at this point: we are not even close to developing something like strong AI, and I am personally doubtful of whether we will ever come to that point.

Published 6-27-16