Yet she [Ada Lovelace] believed that any act of creativity would lie with the coder, not the machine. Is it possible to shift the wight of responsibility more towards the code? The current generation of coders believes it is.
TL;DR: We’re not there yet. But we’re not too far away.
Mind you, this book was published in 2019. Leaps and bounds in the field of AI does not happen by years, they happen by months. Since the publication of the Creativity Code, AI has learned how to:
• Understand jokes
• How to generate an image from a mere abstract suggestion
• Feel pain
There have been seminal moments in the novel but eventful history of AI, even in the last decade that was covered in the book. For du Sautoy, the most pivotal moment in AI was the day when DeepMind managed to beat Go’s world champion, Lee Sedol. In all honesty, I know much about a game of Go as I do about DOTA — It’s there, I watch the contest more than the game itself, but not something that I would likely learn in the future. Go is far more complex than chess and though chess GrandMasters now struggle to beat chessbots, Go masters were still able to beat the computer in Go without as much as breaking a sweat.
I watched the doco on this game and to this day is still my favourite documentary that I visit now and then in YouTube. Lee Sedol, humiliated by a cold hard machine, looking at the friendly face of AJ Wang, who moves the pieces as AlphaGo would if he had a mechanical hand (thankfully we weren’t there yet) was ambushed by AlphaGo’s fluid creativity. I feel for the guy because Go is as big as Gordon Ramsay is in Hell’s Kitchen in South Korea, and the flashes just exacerbated his humiliation in the post-match interview.
To mathematicians, Lee Sedol’s defeat is a sobering moment. For them, mathematics is still something untouched, a discipline which requires a mixture of human creativity, genius and balls of steel in order to solve proofs and equations which in turn, will progress human thought. For once, Du Sautoy felt that his job was in danger. This existential crisis drove him to write the book, to thoroughly research where AI stands as a discipline and to get in touch with AlphaGo’s head honcho, Demis Hassabis who back in the day, was also a chess prodigy.
De Sautoy’s approach is to log the progress of AI in the fields where human creativity is most prominent: in art. AI can now generate music based on Bach’s composition to the point of fooling experts that its creation is better than the original. AI can also can paint as it has been instructed to do, but following the mood of the day. The “Painting Fool” is able to create a painting which is depressing in tone if the news headlines at the epoch is also doom and gloom. AI can also write and generate stories which read better than Twilight or the majority of English majors.
If this worries you, it shouldn’t. Yet. In every single one of these projects, there is too much of the shade of the creator for the AI to be considered fully independent. Perhaps this statement is apt:
Art is ultimately the expression of human free will, and until the computers have their own version of this, art generated by a computer will always be traceable back to a human desire to create.
What we think of machine intelligence, really is our own desire to push our creative boundaries warped by the interpretation of algorithms.
What I appreciate about The Creativity Code is the readability despite the natural complexity of the subject. Du Sautoy briefly touched on the evolution of machine learning and how an algorithm is trained like an athelete in the Soviet Union. When he does, I needed to read the pages a few times, but you’re never lost and it is worth focusing on these because some of the methods use to train algos are fundamental and seemingly so simple. Some of these algos are now used to match medical students to their residencies.
As a novice in the discipline, The Creativity Code is a good introduction of where we are in 2019. I fear that these types of books, as plenty of political science books, will quickly become obsolete as father time vindictively tries to prove naysayers wrong. Du Sautoy’s conclusion is more open-ended though. We are some distance away from technological singularity, but the discipline is travelling in lightning speed to close that distance. I’m not sure I’d like to be there when that happens.
NOTE: This review was published in Goodreads in 2022. Since then, ChatGPT has brewed up a storm. The verdict is still the same: we’re not there yet. The technological singularity is still a few horizons away. But the fact that anybody with an internet connection can access AI as a resource now is already a game changer.
