A good friend of mine, Brian Christian, has recently written a book and a teaser article in The Atlantic about Artificial Intelligence and what it can teach us about our humanity. As a physicist, I enjoy learning about AI and other broad technological advancements, and as a human, I enjoy learning and thinking about humankind; I find the intersection particularly intriguing.
At issue are the defining characteristics that make us uniquely human. Historically, we would compare and contrast ourselves with the rest of the animal kingdom. We would make statements like “Humans are the only animals that use tools” (a fine theory until primates were observed doing the same thing), or “Humans are the only animals to use language” (until discovery of communication in dolphins, whales, and other species). With the advent of computing and advances in AI and machine learning, the list of uniquely human attributes is dwindling rapidly. To start, computers possess memory and arithmetic skills that easily outclass humankind. Perhaps more interestingly, computers have demonstrated superiority in specialized fields like chess (Deep Blue) and Jeopardy (Watson), emerging victorious against the top human contenders – impressive feats given the non-deterministic trajectory of the contests.
So what about us is unique? Is it possible to complete the sentence “Humans are the only animal to __________”, in a way that is accurate now and reasonably accurate in the future? As I ponder this, I can identify two broad areas in which the quintessential essence of humanity shines through (and I’m sure there are many others that I haven’t considered).
The first is one of subtext – an ability to “read between the lines.” We can tell when someone is bored without asking directly. When a friend says “everything’s fine,” we can tell immediately if everything’s fine, or if there is some concealed crisis. Bribery, seduction, and threats (to use the words of Steven Pinker) can all have their desired effects without being spelled out explicitly. As humans, this comes so naturally to us that most people take it for granted.
Subtle body language, wordplay, and innuendo all stem from the root of shared experience. A human from a different culture may easily misinterpret or miss altogether these forms of under-the-radar communication. Having a computer attain fluency in such a medium seems to me reasonably out of reach for now. It’s worth noting that Watson performed most poorly when faced with questions involving puns, jokes, or words used in unusual contexts. Deep Blue played chess most weakly not during the opening (memorizable) or end-game (calculable), but the middle-game, where strategy, subtle positioning, possible gambits, and intent all need to be analyzed.
A second broad area that seems exclusively human is that of imagination and the will to create. The obvious application of imagination and creativity is to the arts; humans hold the monopoly on art for art’s sake, and computers have a long way to go to catch up. This may not be altogether unexpected. The arts are a subset of the humanities – those subjects which seek to inform us about the human condition. It is an enormous challenge for a computer, without any shared human experience, to teach us something about humanity. People certainly try; I like the example of a program that composes in the style of Bach. It can do pretty well for a short while, but then gets stuck in a rut – it fails to surprise us with new keys, transformations, and motives as would be expected from Bach. In a sense, these masters of high art – Bach, Shakespeare, Michelangelo – are guardians of our humanity, in that their creations stand alone as something only humanity can accomplish.
Of course, imagination and creativity are not limited to the arts. I would argue quite strongly that these concepts are of supreme value in the sciences, physics included. We may have ceded the grunt work of equation solving and data analysis to computers, but ideas are human borne. Einstein’s quote – that “imagination is more important than knowledge” – rings prescient here. The capacity of the human mind to see reality as it is, and imagine the unknown, underlying governing laws, seems safe from the encroachment of computers that necessarily rely upon pre-programmed, well-known governing laws.
I believe that identifying unique aspects of our humanity can have far-reaching implications, but I want to restrict myself to a quick discussion of education for now. In the US, modern schooling revolves around standardized testing, in which creativity is too often subordinate to fact memorization. This is troublesome in an era where we are losing (have already lost?) our knowledge supremacy to computers. It seems as though we should grab hold of creativity or whatever else makes us uniquely human, and build a culture around encouraging these pursuits. Otherwise, we may simply be leading future generations down a path of obsolescence.
Whatever the implications, thinking about the relationship of AI and our own humanity can teach us something new about ourselves, which in my mind makes it a fascinating subject.