Here's a job ad you haven't seen but soon will: Biology teacher, high school, certification in robot training. To get smart, robots need to get an education.
Teaching robots runs counter, I realize, to everything you know about the robots in your life. Your new iRobot Roomba 790, with iAdapt behavioral switching and new wireless command center, doesn't need to be taught to use its optical sensors to search and destroy the dust bunnies under the china cabinet. Your Mercedes CL550 Coupe, with PRE-SAFE and Distronic Plus adaptive cruise control, works the first time, out of the box, to sense stopped traffic and avoid impact.
But once in our hands, these intelligent machines don't impress us. As my human students say, "But they don't have robust human-like intelligence." So how do we get that? Just like we do in humans. We teach them.
As a biologist who builds autonomous life-like robots, and who also happens to be a college professor, I can't help but think about the lesson plans we are going to need in the not-too-distant future when we home-school our domestic robots for mundane but critical skills like putting the doo-doo and not the dog into the plastic waste bag.
What better place to start than teaching robots about life itself? So here is your reading list to get you up to speed so that you can be a better -- dare we say -- parent to the new intelligent machine that you will soon welcome into your life.
1. Rolf Pfeifer and Josh Bongard's How the Body Shapes the Way We Think: A New View of Intelligence may well be remembered as the book that got us started raising smart robots. Pfiefer and Bongard transport us from the classic brain- and algorithm-centric world of classical artificial intelligence into the new paradigm of embodied robotics, where interactions of the robot and its world are necessary for intelligence. Continuing to use biology proof of concept, they create developmental robotics. For robots to grow up smart, they need, among other things, the ability to move their bodies and sense their world in a variety of different ways. As the computer control system learns, the body changes its size and shape. In combination, the learning and growing robot can slowly scaffold its intelligence about its own capabilities to an understanding of what the world means in terms of the actions it can undertake to achieve its goals.
2. Louise Barrett's Beyond the Brain schools us on a different set of bio-blinders: our myopic anthropocentricism. Barrett is a psychologist who understands that when we only study humans we isolate ourselves in an abstract box that makes self-understanding difficult. Oh cruel irony! To get us out of this box, she advocates that we look beyond humans, beyond our obsession with ourselves and our enormous brains, in order to put humans into context. That context is evolution and we are evolved animals. As such, we bring with us a host of genetic and physical features that are shared by our close and distant cousins. If you want to understand humans' cognitive psychology, you have to understand the cognitive science of animals and robots modeled on animals.
3. Jon Weiner's The Beak of the Finch does something extraordinary: it tricks us into understanding evolutionary theory by telling us stories about people. I recommend it as a welcome replacement for Charles Darwin's soporific On the Origin of Species. While I also recommend Origin (only the powerful first-edition facsimile, and not the usually-available and watered-down sixth), it's a book that requires, even for biologists and probably robots, repeated administration of strong coffee. Weiner's Finch, on the other hand, is itself a stimulant. Weiner enlivens the usual suspects, like Darwin, and introduces us to the modern biologists Rosemary and Peter Grant. Grant and Grant observe and measure and explain natural selection in action, as it happens. Weiner, the embedded reporter, tells us first-hand what the Grants are up to in the Galapagos Island. In dry years, with few seeds available, big-beaked finches outlast and out-reproduce those birds with smaller beaks. In response, the next generation on average has bigger beaks. That's evolution. As an environment changes, so too does the population of life forms there that must hourly, daily struggle for existence.
4. Debra Gordon's Ant Encounters: Interaction Networks and Colony Behavior is arguably the best account of the behavior of individuals of any species and the consequences of their behavior on their societies. Gordon, Professor of Biology at Stanford, and her students have spent years watching individual ants at work. She is another destroyer of paradigms, laying to waste the anthropocentric misconceptions still used by engineers to build intelligent "ant-like" algorithms and swarms of robots. Ants aren't born into castes. Instead, individuals transition from work inside the nest to work outside. What signals an individual to start work each day or to switch jobs depends on the collective behavior of the networked colony as it is reflected in the behavior of other individuals. Moreover, the colony is not a monarchy, with the queen in charge. Instead the colony is super-organism acting with cell-like coordination to reproduce, forage, and defend itself.
5. In a fit of immodesty, let me offer John Long's (yes, that's me) Darwin's Devices: What Evolving Robots Can Teach Us About the History of Life and the Future of Technology. In a twist, Long uses robots to study animals and their evolution. The trick is to build robots that are models of animals equipped with essential bio-realistic features like the ability to behave autonomously, without any human pulling remote control strings. Because these biorobotic models are simplifications of real animals, their behavior and evolution is easier to understand. What robots teach us is that both randomness -- in the form of mutations -- and structured probability -- in the form of natural selection -- are essential for the automatic, creative, and surprise-producing design process that we call evolution.