While I believe that creativity will be perfectly explained one day. I also believe it will be far from being easy to be replicated artificially. I’m basing such a sweeping proclaim on a simple observation. That of AI and the way it has been implemented in chess programs up to now.
In fact, although nowadays the best chess programs are so strong they can often beat even the strongest Grand Masters in the world—it all started with Deep Blue and Garry Kasparov’s match in 1997—these programs play in a way that is quite different from the style of any human chess player.
In fact, microprocessors, because of the way they are designed, are more easily programmable along certain principles rather than others.
For an interesting article about the difference between computer and human mind go and read this Lance Whitney‘s article.
A computer can spit out in rapid succession a list of prime numbers. Or tell you the day of the week one million years two months and three days from now. But the moment you tell it a joke any six year old would instantly get, it freezes.
Of course the march of progress seems unstoppable, and lately I’ve read about a new program, AlphaZero, that seems to learn like humans do. That’s to say when it tries to implement new ideas in its style, for a while it seems to be weaker, as if it had to find the best way to embed the new knowledge into the bedrock of what it already knows.
This is in itself a notable achievement. However whether or not such a learning style leads also to a more human-like style when playing a game has still to be seen. After all, the way humans learn has not direct bearing on the way they consequently put that learning to use.
A matter of degree or a matter of kind?
Chess is an extremely complex game. This is pacific. But for all its complexity it has nothing of the complexity of creative writing.
Just think about it.
Chess has a relatively small set of different pieces, 32 to be precise. There are two players and they can make only a restricted set of legal moves. Such moves are limited to the sixty-four squares composing the chessboard. Lastly, the objective for each player is clear and always the same: checkmate the enemy king.
Instead, creative writing offers something like 150.000 words to choose from as pieces. Such words can be reused over and over, at will. Literary works can range in length from few hundreds to millions of words. The players, the people involved, are innumerable. In fact the writer, usually one, is complemented by a number of potential readers that can exceed the millions. Finally there’s no one simple goal. Not for writers and not for readers.
I could go on for pages and pages, but it’s already apparent that if in a game where the rules are few and unambiguous the best programs still play in a way that gives away their non-human nature, then in creative writing, where rules abound and are ambiguous and shifting at best, the chances for an algorithm, or a cluster of algorithms, to manage all those variables like humans do are next to zero.
It’s a bit like the old story about all those monkeys typing at random and coming up with all the works by Shakespeare. Sure. That is possible in principle. If you happen to have infinite time and patience and zero ideas about how to better spend your time than waiting for the damn monkeys.
In fact an essential difference between chess and literature is that while in chess a program can beat a human opponent playing like a robot, in creative writing there’s no way to come up with a creative work without adopting a human style–I would be tempted to say: without becoming human.
Of course, unless we want to breed a new race of synthetic life, perfectly suited to understand and laugh at jokes we humans will never get.
123456 and 7
234567 and 8
345678 and 24… Ops… Sorry I’m dyslexic. I keep swapping addends with results.
This is just a lame example I came up with. It doesn’t want to be rigorous. After all I’m pretty sure I don’t have a microprocessor inside my head, so I can only try to imagine something abstract enough to prevent us humans from getting it at once, but also explicable enough to allow us to understand the
terribly lame joke once it is explained.
A bridge can be built by AI and if it is stable it can be considered a success. The same goes for many other products. Roads, tunnels, electric circuits, and programs. Indeed, in the case of programs, having them written by other programs could even improve the robustness of their code.
However, for something still so elusive and changing as creative writing, for novels and poems, I believe we’ll have to wait for a good while.
This can be a good thing or a bad one. For sure a silicon-powered writer could write one hundred of books in one night. So fans of a series would never run out of new stories.
But then if any story can be written any moment, than all stories would lose their value. They would turn from precious stones into anonymous grains of sand.
I don’t know about you. But a world where books are as common as sand and just as ordinary it’s a world I’m not sure I like. Not in the least.