Why the Next Frontier of AI Might Be Child’s Play
At the peak of the knowledge economy, the premium we placed on intellect could be clearly measured across the market. Firms invested heavily in opportunities to recruit the best and brightest, with the cost of acquiring this human capital justified by measurable returns to shareholders. Then, something shifted.
Capital began circumventing our grey matter and was instead injected into AI, leading us to question: What happens when the value proposition posed by the modern human brain – which has been developing for over 200,000 years – begins to change?
The Turing Test
In the early 1950s, computer scientist and mathematician Alan Turing posed a new question: Can machines think? This gave rise to what has become known as the Turing test, which assesses a machine’s ability to exhibit intelligent, human-like thought and behavior. One way we have tested machines in this regard has been to have them compete against humans in games. In other words, we have asked machines to exhibit their humanness by asking them to play, to very successful results: IBM Watson, for example, defeated two reigning Jeopardy! champions, while Deep Blue’s 1997 victory against chess champion Garry Kasparov signaled an important shift in AI.
However, while these successes showed that machines could logically outsmart humans, even the strongest proponents of this technological advancement hesitated to conclude that we had truly recreated the ability to think like a human. Even now, the concept of distilling hundreds of thousands of years of human knowledge into a few microprocessors seems like an insult to the complexity that is the human mind – a complexity that we have yet to fully understand. Indeed, if given a task that relies on imagination, machines would most certainly fail the Turing test. After all, imagination is a uniquely human trait.
The Evolution of Imagination
From an evolutionary biology perspective, our ancestors’ development of imagination was a significant change that led to our species’ marked departure from chimpanzees. While other animals dedicated resources to developing more powerful jaws, thicker scales, longer claws, and other useful traits, the evolution of Homo sapiens lent itself to increased brainpower. Creativity manifested itself as a key adaptation for the development and survival of our species.
Unlike other mammals, Homo sapiens made and used tools, and harnessed fire to cook. This innovative procurement of food led to the establishment of highly coordinated hunting systems and societies, which eventually grew to become states and nations. Over time, we devised ways of governance by recording scripture and law. We also constructed societal norms, traditions, and ideologies – all of which came from our imaginations.
For a human brain to fully develop, two main components are required: energy, in the form of large amounts of calories, and a period of prolonged growing. Childhood, a state that begins as a drain on communal resources, ends up paying dividends later, as it allows the human brain to build the neuronal ensemble scaffolding critical to imagination.
How Childhood Builds Innovators
During childhood, the brain exhibits strong plasticity and a capacity for varied myelination leading to synchronous firing, a key component of imagination. Cumulative cultural evolution describes another uniquely human phenomenon: the propensity for copying behavior. Children exhibit this habit throughout their development when they play pretend, which allows them to situate themselves in imaginary scenarios that mimic those navigated by adults. This imitation shapes the developing brain’s wiring and firing, allowing children to quickly acquire skills. The ability to learn through imaginative play is a competitive advantage for our species, as even into adulthood, we are able to creatively build off of the foundational knowledge we acquire in childhood. In short, the pretend play of childhood serves as ground zero for the imagination that society itself is built upon.
With no other species’ offspring existing in such a dependent state for so long, humans are able to reach economies of intellectual scale unmatched in the animal kingdom. But is it possible that this advantage has run its course in the emerging man–machine intellectual hierarchy?
For decades, humans have worked to advance the capabilities of machines. By evolving and programming the “genetic” code of computers, we have advanced technology, reduced our own cognitive loads, and subsequently gained the freedom to imagine the next application of our innovations. Recently, we took a quantum leap and allowed machines to learn and to teach each other. In other words, we programmed AI – and it, in turn, began programming itself.
In 2011, AI took a new step toward truly passing the Turing test with the birth of a sub-discipline of machine learning known as deep learning. Deep-learning algorithms seek to truly replicate the way the human brain works by building connections between synapses. Most recently, Google’s DeepMind began inputting imagination-augmented agents (I2As) into its software, which helps the AI sort valuable predictions about its environment from predictions with little value. The era of synthetic imagination is upon us.
The most compelling part of this new wave of intelligence is how this synthetic imagination learns. As was done with IBM Watson, researchers continue trying to bolster the ability of AI to out-compete the human mind by asking it to play. Until recently, games where machines must imagine possible scenarios and build strategy had puzzled programmers. Then, in 2015, AlphaGo defeated champion Lee Sedol at Go, an ancient Chinese strategy game with exponentially more possible scenarios to imagine than a logic-based game of chess.
The first iteration of AlphaGo was impressive, to say the least. However, the speed of what happened next was startling even to researchers. While AlphaGo learned by studying the moves of the most skilled human player of Go, the next iteration, AlphaGo Zero, had no such training. AlphaGo Zero was programmed only with the rules of the game. It then played itself over and over again, learning and improving with each game. In just 40 days, AlphaGo Zero became the greatest Go player in the world – and it did this all through self-play. In a way, the AI underwent an artificially programmed period of childhood, which allowed for a wiring and rewiring of artificial neural networks through pretend play.
The Next Generation of Imagination
The feeling of liminality associated with this point in history is palpable; humans are no longer assured of the value of the human mind in tomorrow’s economy. On paper, it appears that we humans have started to remove ourselves entirely from the equation – the right hand no longer knows what the left hand is doing. But before we arm ourselves for a neo-Luddite revolt, we should stop to reflect on this alarmist narrative.
Aided by technology, humans now imagine in ways we never could before. We’ve imagined a world in which we no longer control the wheel of a car; in which space tourism is the norm; and even in which 3D-printed organs extend our mortality. And it was humans who imagined the very machines that now challenge our standing as the sole creators of new ideas.
The human mind is no longer the only key player, as it is increasingly challenged by machines – but these challenges can be the very thing that pushes us to evolve into the most exciting period of human innovation yet.