46 Can We Apply the Hallmarks of Creativity to Computers?
We’ve seen that computers can already be said to exhibit a form of introspection, mulling over data as they work on a problem. They also focus, persevere, and are not afraid to make mistakes—qualities built into their systems. But what of the other hallmarks of creativity and marks of genius?
The Need to Know Your Strengths
For us, knowing our strengths is essential to prevent wasting our time on intellectual pursuits for which we are not suited. Einstein focused on physics despite his interest in mathematics, realizing he did not have the sense for what was a fundamental problem in it, and Heisenberg opted for physics rather than a career as a concert pianist.
Computers have the whole world of knowledge at their disposal through scanning the web and therefore have absolutely no limits. Embodiment—when computers can be installed as the brains of robots—will enable them to accumulate emotional experiences, too.
The Need to Beg, Borrow, or Steal Great Ideas, and the Need for Collaboration and Competition
AlphaGo Zero defeated AlphaGo by one hundred games to nil, while AlphaZero defeated all extant chess programs. This was a battle of programs in which the computers reacted automatically to each other.
But what if we built computers that wanted to outdo each other? Suppose a computer called A was composing a piece of music and became aware of another, B, doing something similar, perhaps as a result of communication between the machines or because one published its early efforts on a blog. A might see this as a game that it wanted to win and might steal ideas from B’s preliminary compositions to advance the state of its own, just as we might. It’s a Darwinian world out there, for us and for machines.
This is not as theoretical as it sounds. Scientists have already carried out a competition like this, setting robots against each other, in the laboratory. The robots are tiny s-bot robots that move on two treads, with a six-inch-diameter structure like a hockey puck on top that holds devices for visually signaling to other robots with light. They are controlled with neural networks.
Groups of robots set about a foraging task and exchange colored signals depending on whether they encounter “food” or “poison.” The internal structure of their neural networks forms their genomes, which quickly evolve over five hundred generations in a population of one thousand robots, depending on how successful they are in gathering food. The most successful groups evolve deceptive communication strategies to ward off unrelated robots—robots with different genomes—thereby enabling their own colonies to continue to evolve. Even among robots, there is survival of the fittest.93
The Need to Focus and Not Be Afraid to Make Mistakes
Computers are not afraid to make mistakes. They don’t become discouraged. They just start over again and again. For example, Simon Colton’s The Painting Fool assesses its work and begins again if dissatisfied with the portrait it has painted.
The Need to Thrive on Ambiguity and the Need for Experience and Suffering
As we have discussed, these qualities are yet to come. But there are already researchers like Rosalind Picard working on developing computers that can empathize with human feelings and even have feelings themselves.
The Ability to Discover the Key Problem and to Spot Connections
These are two marks of genius I identified earlier. But how do they apply to computers? Computers think, but in a way very different from how we think. An artificial neural network understands the world in terms of numbers, which is how it encodes incoming information. Working in a particular field such as physics, it can scan every research paper and spot a flaw or a gap in the current attempts at problem solving. If it then turns to the problem itself, it may conclude that this problem is not as fundamental as it first appeared.
The computer might parse out the problem, focusing on a particular aspect just as, in Einstein’s day, scientists sought a theory of the electron. Only Einstein realized that they were all working on the wrong problem. They had still to elucidate the basic concepts of physics—the nature of space and time. To confront this entirely different problem, Einstein had to tap into branches of physics that apparently had nothing to do with space and time—such as thermodynamics, in which the basic statements have to be accepted without proof—to come up with his theory of relativity.
Computers will have to make the same leaps of judgement. After identifying a new problem, the computer mulls it over—employing unconscious thought—running through different possibilities until illumination strikes. Artificial neural networks employ information encoded in numbers, which puts them in the perfect position to do so. They can home in on similarities between disciplines and seek out connections between, for example, unrelated works of art, as exemplified in Mario Klingemann’s X Degrees of Separation.
Computers can work in the same way in the fields of physics, art, literature, or music. All these are problem-oriented and have rules for proceeding. Sometimes artists break the rules, such as when Picasso depicted all perspectives simultaneously, when writers use words to make pictures in concrete poetry, and in twelve-tone music and the atonal works of Stockhausen and Xenakis. If they are ever to be deemed truly creative, computers will have to do so too.
Ahmed Elgammal is already working on this with his creative adversarial network, primed to invent new styles and break with the past. Future computers might scan the scores of many genres of music, identifying similarities and differences, and then evolve a whole new genre. For this, an AI system that mixes artificial neural networks with symbolic artificial intelligence along the lines of a newly structured Flow Machine might create intriguing variations on existing styles.
A whole new future is opening before us: not one to fear, but one to look forward to with anticipation, in which machines work together with us to enrich our lives with new forms of art, literature, music, and much else.