Click to learn more about author Biju Dominic.
When AlphaGo made the 37th move in the second game of the famous match between AlphaGo and the leading Go player Lee Sedol, all the commentators of the match were shocked. It was a move no human player would ever think of making. In the thousand-year history of the game, there is a rule of the thumb that even beginners of the game know – avoid placing stones on the fifth line from the edge. But that is exactly what AlphaGo did. It turned out to be a significant move that helped AlphaGo win that match. That marked a watershed for many AI enthusiasts: In their view, AI had moved past human influence and cognitive ability and was able to stand on its own.
If this is accurate, why has the progress in tackling next-level AI challenges been so elusive for machines? Because it’s not about competition, but cooperation. AI and neuroscience need to learn from one another.
AI has seen great progress, especially in supervised learning. It is now boldly taking on several new challenges, such as managing unsupervised learning tasks, tackling one-shot learning, and even acquiring general intelligence. Yet, significant hurdles remain. And there’s a tool that already exists that has effectively executed these tasks and provides a perfect guidebook for the future of AI: the human brain.
Here are a few reasons why the human brain and neuroscience are key to the future of AI.
Why the Human Brain?
The brain is still the most intelligent, energy-efficient machine in the universe. In fact, among the intractable problems AI is targeting, nature has already solved most with our own brains. In his Pulitzer Prize-winning book “Gödel, Escher, Bach: An Eternal Golden Braid,” Douglas R Hofstadter wrote, “All intelligence are just variations on a single theme; to create true intelligence, AI workers will have to keep pushing closer and closer to brain mechanisms, if they wish their machines to attain the capabilities which we have.” So, the more we understand how the brain accomplishes a task, the more we can shorten the learning curve for machines.
The idea that machines should take the lead from humans is not a new one. For example, a hydraulic crane didn’t just learn how to lift things more efficiently on its own. The radio telescope didn’t just magically happen to see farther than the human eye. Instead, we built these more powerful tools by taking inspiration from how humans performed those tasks. There is much more to be learned.
How AI Can Use the Human Brain
Beyond the inspiration and insight it offers, the human brain’s performance acts as a validator and helps avoid reinventing the wheel as we design the machines of the future. Take visual recognition, for example. Our visual sense has evolved over millions of years, and the study of the brain’s visual cortex has inspired some of the most successful deep learning algorithms.
Neuroscience can also inspire practical benefits. One amazing fact about the human brain is its extremely low energy consumption. It performs by using just 20 watts of energy. On the other hand, a supercomputer – which is not as powerful as a human brain – consumes more than five megawatts, or 250,000 times more power than our brain. DeepMind, the creator of AlphaGo, is reportedly using neuroscience to help Google reduce its data center’s cooling bill by 40%. We have only just begun to tap the value of the science of the brain.
Establishing Symbiosis Between AI and Neuroscience
In the last two decades much new knowledge has emerged about the human brain’s functioning. The emerging fields of system neuroscience (the study of how different components of the central nervous system interact to produce experience and generate behavior) and computational neuroscience (where mathematical tools and theories are used to investigate brain function at an individual neuron level) have added great value for AI professionals. As V. Srinivasa Chakravarthy, author of “Demystifying the Brain: A Computational Approach,” wrote, “These new fields have helped unearth the fundamental principles of brain function. It has given us the right metaphor, a precise and appropriate mathematical language which can describe the brain’s operation.”
At the same time, it is also true that neuroscience can benefit from AI. The study of the human brain can benefit tremendously from AI’s power. For example, mapping the hundreds of trillions of synaptic connections in the brain is estimated to generate zettabytes of data. (A single zettabyte is one million petabytes.) And this is the data from just one human brain! So if we are ever to unravel the complexities of the brain, AI will have a major role. This means that in addition to introducing neuroscience into the AI equation, we will need to create a new environment where both fields of study work together.
Instead of walking alone or having an adversarial relationship with one another, AI and neuroscience must join forces. Many of AI’s earliest innovators – Terrence Sejnowski, Geoffrey Hinton, and David Rumelhart, for example – had strong backgrounds in neuroscience. The more AI studies and understands the inscrutable algorithms of the human brain, the better our technology will be. Magic can surely happen if we achieve this marriage of the mind and the machine.