Fans of artificial intelligence (AI), such as Google’s Ian Goodfellow, the inventor of the phenomenal generative adversarial network, argue that “machines are already creative” in the book The Artist in the Machine: The World of AI—by Arthur I. Powered Creativity. Miller shows many examples of AI-generated art, music and literature. Is this proof enough that computational creativity has reached human-level creativity?
Margaret A. Bowden, Professor of Cognitive Science at the University of Sussex and author of the book, The Creative Mind: Myths and Mechanisms, suggested three criteria for assessing whether an idea or artwork is a product of creativity: it should be ‘new, surprising and’ needed. Precious’. Christie’s recent auction of an AI artwork called the Portrait of Edmund de Bellamy for $432,500 provides clear evidence that AI creativity has many buyers. But the bigger question is whether AI processes can really churn out amazing new ideas.
Since 1908, when the French mathematician and philosopher Henri Poincare came up with his four-stage framework for developing a creative idea, it has remained the benchmark of an ideal creative process. The first stage of the creative process involves feeding the brain with all the information it will need. The second step is trying to find connections and contradictions between all the collected information. The third stage is where the problem solver is expected to ‘get out of the problem’ and engage in other activities that are relaxing. It is during this stage of relaxation that a creative idea is expected to leap from nowhere. Then comes the fourth stage, where the rough edges of the creative idea are polished to make it perfect.
There is no doubt that with the vast level of information available on the World Wide Web, computers will be able to collect much more information than the human brain can collect. With their enormous computing power, today’s computers are far more efficient than the human brain at identifying relationships between different pieces of information. It is possible for a computer to take care of evaluating a creative idea and drive it to completion. The question is whether digital machines can effectively do the third step of ‘getting out of the problem’. Many experts in creativity would argue that what happens in the third stage most likely determines whether the output will be a general idea or a brilliantly creative one. In this stage of the creative process, someone is taking the problem to an unconscious level and using their tremendous efficiency to generate connections like never before in the human brain. These are completely unpredictable until new connections are made. This is exactly what makes the creative idea so amazing. There is also a mythological term “eureka” of the joy experienced upon such discovery, as said by Archimedes in defiance of the condition of his clothes. But can AI really ‘get out of the problem’ as can humans? Can AI replicate the smart non-conscious processes of the human brain?
A team from DeepMind developed AlphaGo, a computer program that can play Go, considered the world’s most complex board game. AlphaGo easily defeated best human go player Lee Se-dol. A year later, DeepMind released an improved version, AlphaGo Zero. It was only taught the basic rules of the game. The rest he learned by playing against himself a million times. In three days, it beat AlphaGo from 100 to zero. Today, there are many who believe that the hidden layers of software’s deep neural networks that explain its enormous self-learning potential are on par with the vast and efficient non-conscious processes of the human brain.
If so, AI should be able to extract creative ideas in great numbers. Where does computational creativity go from here? Could AI be the harbinger of world-changing innovations?
Innovation is essentially about taking newly created ideas and developing them into something useful and practical. But turning creative ideas into innovation is no easy task. In its early stages, it is very difficult to separate a good creative idea from many other generic ideas. It is like identifying the future success among the many newborns in the pediatric ward. The history of great ideas reminds us that sometimes even the innovators themselves do not realize the potential of their ideas for many years. For 10 years after his famous journey, the idea of evolution remained in the notebooks of Charles Darwin. It was only after Alfred Russell Wallace, another biologist, reached the same conclusions as Darwin, did he really begin to believe in the true power of his proposal.
In the book, Where Good Ideas Come From: The Natural History of Innovation, Steven Johnson reminds us that to build a world-changing innovation, just one brilliant idea is not enough. Many other things in the environment have to fall into place. For example, the idea of a video-sharing platform like YouTube didn’t take off until millions of people had access to the Internet via a reasonably fast dial-up connection. The idea would have been considered ‘ahead of its time’ if it had been tried before circumstances suited it.
AI should be able to generate many new and surprising ideas. But facilitating an efficient transition of these creative ideas to true innovations would require a great deal of human intervention. It is clear that the future of computational innovation will be a lot about strong AI-human collaboration.
Biju Dominic is Chief Promoter, Fractal Analytics and President of FinalMile Consulting.
catch all business News, market news, today’s fresh news events and breaking news Updates on Live Mint. download mint news app To get daily market updates.