Do Androids Paint Electric Sheep?
By Arijit Banerjee
An enquiry into the potential for consciousness of Artificial Intelligence
As the twenty-first century progresses and Artificial Intelligence replaces more and more human professions, an increasingly relevant question is: what can humans do that AI cannot?
The answer, only a few years ago, would have been “the creative arts”. Bots and algorithms do not possess the hearts and souls that drive humans to make art. How would the creation of a mindless machine be able to touch us the way that a Van Gogh painting does? How would the searing emotions we feel while listening to a Beethoven symphony be replicated by a silicon chip?
All these assertions have been turned on their head in recent years. As AI grows more sophisticated and machine learning algorithms get ever more complex, it seems that there is nothing that it can’t do.
In Luxembourg, an AI known as AIVA has been trained to compose music by processing databases of classical composition. AIVA now composes music for video games and has been given copyrights. In the US, an AI learned from a collection of paintings to generate its own, and scored more favourably with critics, compared to human artists.
If machines can so easily do something that we humans have always considered unique to us alone, then how are we unique at all? What sets our consciousness apart from that of a machine? Does a painting and music-composing AI deserve to be considered similar to a human?
What is Consciousness and can AI develop it?
The once widely-accepted idea that human consciousness is unique is discredited by the latest work in cognitive neuroscience. Researchers have found that the human brain behaves like a sophisticated prediction engine., i.e. it perceives what it expects to perceive, though sensory information does rein in some of the wilder predictions. Consciousness is the brain hallucinating its expectation of reality.
Similarly, “selfhood”, like consciousness, is a complex set of brain-generated constructs which converge to our experience of ourselves as individuals. These include:
- The bodily self, i.e. “I am this body.”
- The perspectival self, i.e. “I see the world from the first-person point of view.”
- The volitional self, i.e. “I want to/need to do this.”
- The “narrative” self, i.e. “I am an individual built of unique personal memories and memories.”
- The “social” self, i.e. the construct of “I” as we think others perceive us.
So how is human consciousness different from other living beings? After all, animals appear to be motivated only by instinct. Experimental evidence, however, points to complex social behaviour and long-term memory in most animals, ranging from chickens to elephants. This new perspective on consciousness then shows that we are not fundamentally different from animals – we merely differ on how “conscious” we are. Evolutionarily, consciousness developed because it allowed us organic beings to look out for ourselves, for each other, and to successfully reproduce.
This leads to a simple conclusion: the experience of being alive is a result of our organic biology. Machines simply cannot, by definition, be conscious or “hallucinate” themselves into existence, into “selfhood”, the way that we do. A machine is not going to wake up tomorrow, conclude “Cogito, Ergo Sum!” and become human.
The Appeal of Machine-Made Art
Let us imagine that we are standing in the ruins of the Parthenon in Athens. Our eyes appreciate the symmetry of its structure, the pleasing straight lines of its columns. What we do not consciously know is that the ancient Greeks, having some idea of the mathematics of perspective, carved curved lines that our brain interprets as straight, in pursuit of underlying mathematical harmony.
Figure 1 (from left to right): The Parthenon in Athens, Sheikh Lotfollah Mosque in Isfahan, and Raphael's School of Athens .
Islamic art uses elaborate geometric motifs; the artists of the Renaissance used complex ratios and calculations to create visually pleasing works; the great symphonies of Beethoven follow the rules of symmetry. We tap our feet unconsciously to a catchy tune, beating out a rhythm.
The pleasure we derive from art is two-fold. It is partially due to the subconscious appeal of the mathematics that we cannot see, a result of the way our organic “consciousness” has evolved. The rest of it is because of the emotional resonance we feel with the artwork itself. This emotional resonance is because of our Narrative Self: we identify with the work of art because it connects to a memory that is part of our construction of ourselves as an individual.
So what happens when an Artificial Intelligence makes art? It is not replicating human selfhood to create it. A machine does not express its raw emotional pain in a masterpiece of pure creation like Beethoven did. It does not have a hallucinated concept of “the self” that it chooses to express artistically. It merely understands the underlying mathematical rules of human art, develops an algorithm, and expresses it. We are the ones who derive emotional satisfaction from it because we are the ones who have an idea of “self”.
A machine might be able to perform trillions of operations in a second. In the future, that might change to quadrillions. Quintillions. This intelligence is purely mathematical, logical. It is incorrect to assume that such intelligence implies consciousness, or vice versa.
We can think of machine-generated art as a sort of creative Turing Test. If we were to look at two pieces of art, one made by a machine and the other made by a human, we might not be able to tell them apart. But when a machine does pass the Turing test, it simply means that it can fool us into thinking that it’s human, not that it’s actually human. The Turing Test was designed to gauge machine intelligence. That is not the same as machine consciousness.
Everyone has come across the Captcha verification online that’s meant to make sure that they’re not an AI but as AIs get more intelligent, they can crack these more easily. Amazon, amusingly, aims to replace these with Turing tests that humans are meant to fail, thanks to the way our consciousness differs from pure machine intelligence.
An Artificial Intelligence can never be more than an imitation, an expression of how we see the world. A Twitter bot becomes racist not because it is an evil AI but because it is taught to be so by observing us. AIl are our creations, but they cannot break the final barrier to being their own mentally-constructed selves. They can only be pale imitations of our ”selves” – a simulacrum of a hallucination, created by us based on our own simplistic understanding of what makes us who we are. For now, at least, humanity is safe from being completely replaced. Human jobs, however, are another matter.
Art of the Future
A more relevant question than “will machines become self-aware in the near future?” is “how will humans use these machines?”
Imagine an art museum a few decades from now. Every single painting, every sculpture, and even the buildings may have been designed by an AI. Human minds might not have had anything to do with it aside from the programming but that’s not going to prevent us from appreciating the art. Our brains will still gaze rapturously upon the hidden mathematics in the galleries and we will still be touched to the core of our beings by the artificially-composed music wafting from the speakers of our self-driven cars. Emotionally, as conscious beings, it will not matter to us that all this was made by machines, not humans.
Consider AIVA again – the company that developed it can generate excellent music for video games, expressing the themes of the game, with very little effort, and considerable profit. Now imagine entire industries that fuse AI with creativity. The possibilities are endless, and even the most prodigious mind in science fiction cannot fully comprehend or describe the consequences on societies and economies.
Fear it or love it, the AI revolution is coming. It will completely change the way we think about jobs, even those, like the creative arts, which we once thought could only be done by humans. Perhaps one day androids will dream of electric sheep, but for now, we do not need to be afraid if they paint them.
About the Author
An engineer by training, Anirudh holds a degree in Electronics and Instrumentation with a minor in Philosophy, Economics, and Politics from BITS Pilani (Goa). He currently works as a Research Associate at The Takshashila Institution, a Bangalore-based public policy think-tank. Anirudh's research covers a wide array of policy questions related to emerging technologies, ranging from social media and democracy to gene editing, network theory, and geopolitics.