As humans, we have always tended to think of ourselves as exceptional. Our ability to reason deeply, our capacity to send and receive vast quantities of information through speech, and especially our potential for self-awareness have all rightly given us a sense of superiority over other creatures. But the rise of Artificial Intelligence (AI) has somewhat changed these calculations, as machines have surpassed humans in many tasks (e.g., in chess). The significance of the AI revolution could be staggering: Might not the importance of our affairs, so predicated on our exceptionalism as a species, diminish if we don’t even retain our place as the finest exemplars of thought and reason?
Consider creativity, the conventional beacon of hope for humans in the battle for superiority over AI. While we have ceded much ground to AI in memory and computation, academics such as Memorial University sociology professor Anton Oleinik argue that creativity lies outside the reach of machine intelligence. Something about creativity’s demand that one shatter the norms of an artistic field, says Oleinik, is incompatible with AI’s tendency merely to master the data it is given and to produce variations on this data. At first glance, recent developments in AI music composition would seem to disprove Oleinik’s theory: The Artificial Intelligence Virtual Artist (AIVA) has been flaunting its recent AI-driven compositions, which do not sound entirely devoid of style or form. But by examining more typical examples of AI’s musical output, the popular classical-music-analysis channel Inside the Score concludes, in line with Oleinik, that music composed so far by mechanical intelligence is dry, unintelligent, incoherent, and thoroughly uncreative. As for the famous examples of AI compositions, Inside the Score contends that these are either doctored or produced by sheer luck after thousands of poor attempts — monkeys and typewriters! Neither possibility speaks well of AI’s creative capacities.
There are several reasons why AI has yet to match humans in creative endeavors. Unlike in more quantifiable fields, humans must necessarily judge the success of AI at producing art in a slow and subjective way. This makes for inaccurate and inefficient feedback to the machine’s algorithms. And while machines thrive on parameters and instructions, art is open-ended and infinite in its potential. But there is a deeper reason for AI’s lack of creativity: It lacks true understanding, a necessary requisite for the emotion and passion present in great works of art.
In 1980, philosopher John Searle crafted a thought experiment in his essay Minds, Brains, and Programs that aimed to disprove the hypothesis that machines could truly have understanding. Searle imagines himself in a locked room, with men standing outside the door and feeding him a story in Chinese characters through a slot. The men receive Searle’s “questions” about their story (also in Chinese) through the same slot and conclude that he must understand the Chinese language. However, unbeknownst to the men, Searle has a large sheet in front of him on which instructions are printed for receiving certain combinations of characters and returning others in response. Searle is mindlessly receiving sequences of characters, following his sheet’s instructions, and returning other sequences of characters. He need not understand the story, or even know that the characters are Chinese, to feign comprehension. In representing a computer through his parable, Searle shows that a program-running machine does not really understand — it only shuffles symbols mechanically.
In the same way, AI lacks the understanding of what it means for sound to be emotive. There is no concise value that can be assigned to the passionate ardor one feels at the opening of Mendelssohn’s Piano Trio No. 1, nor to the tristesse one undergoes in the Funeral March of Chopin’s Piano Sonata No. 2. And even if we could train AI to produce motifs that sound pleasant and authentic — as we have started to — the stretch from short phrases to a long work imbued with an underlying artistic message is vast. There seems to be something unquantifiable and ineffable about the artistic license displayed by the greatest composers, something we can understand only by virtue of having a mind that feels emotion and grasps sound beyond the mechanical plane.
Yet we have been wrong about AI before. It is not unimaginable that, through the brute force of its computing power, machine intelligence will find a way to surpass humans someday even in creativity. Still, our fierce pride would be left with some consolation.
First, a creative AI would still leave us as the only understanding beings on Earth. Devoid of a neural network, AI does not have consciousness or reason — it is a mindless if incredibly efficient machine.
Second, even if we manage to construct a neural network for a machine, and this machine can think in a way a computer running programs cannot, we can still lay claim to the glory of causal precedence. When we invented the automobile, we did not lose our minds at being surpassed so dramatically in speed, but rather reveled in our ingenuity and craftsmanship at having created this apparatus. In the same way, even if we are surpassed by machines both in creativity and understanding, we can take comfort in the fact that they would not exist if not for our exceptional ability to innovate. We are the true creative beings, and no machine can take this away from us — even if it out-composes Bach one day. So let’s go ahead and make a toast to our robotic friends: “You wouldn’t exist if not for us, derivative ones! But please, please, don’t become sentient.”