The Creativity Code—Marcus du Sautoy

16 May

For me, it’s all about socks. I’m obsessive about wearing a matching pair—unlike my eldest son who just grabs the first two in the drawer. I find that behaviour even odder than his socks but, then, he finds my sock-matching fetish equally peculiar.

Obsessions can, of course, be debilitating. My socks are in various shades of washing machine-faded black and precision-pairing is time consuming. But, one wonderful day, a fabulous time-saving thought came to me in a flash; if I can’t see the difference then it actually doesn’t matter! Now I just grab the first two darkish items in the sock drawer and put them on. Like son like father.

This marvellous new way of seeing the world was truly liberating. Corn Flakes need not be Kellogg’s and my British gas is no longer supplied by British Gas. My next vacuum cleaner purchase could be something other than a Hoover whilst a broad vista of yeast-extract options has opened out before me.

Marcus du Sautoy. Photo © Fernando Bagué

And then, on Wednesday night, along came Marcus du Sautoy with his deep thinking on deep learning; more specifically the use of deep-learning programming as a way to make computers creative. Absurd? The creations of Artificial Intelligence (AI) would be, well, artificial. But—with my new found sock-wisdom—if you can’t tell the difference, it doesn’t matter. Does it?

In the field of AI, the idea that distinctions without a difference don’t matter goes back to the inventor of computers, Alan Turing (yes, we can debate who actually made the first computer but Turing invented the ‘Turing Machine’ concept they are all based on). Turing was fascinated by the idea that, one day, computers might have real, human-like intelligence and he suggested that we might test a machine for intelligence by holding a conversation with it (possibly through a computer screen). If you can’t tell that you’re talking to a computer then, according to the Turing Test, it’s intelligent.

Unfortunately the Turing Test has, itself, failed a test—the test of time. Programs that successfully fool humans this way have been written, but they cheat. They use tricks to reflect back to people what they have themselves said and they do this in a convincingly human-like way. But no one thinks these programs are intelligent in any meaningful sense.

A new approach was needed and so, in 2014, a team of computer scientists led by Selmer Bringsjord came up with a new IQ test for computers—the Lovelace Test. Named after Ada Lovelace (the 19th century, female inventor of computer programming), this requires a program to repeatedly create something new—a picture or poem for example—and it must do it in a way that is not transparent to those who wrote the program. This last part is important. I can easily write a program that produces an image never before seen, but precise instructions on how to produce that picture would have been put there by me and the creation would be mine, not the program’s. A truly creative program needs to conjure the image out of thin air, just like human’s do.

Marcus du Sautoy’s talk showed us examples of programs that, arguably, did precisely that. Pictures and poems split the audience fifty/fifty on whether they were made by humans or by AI. I did quite well though. My rule of thumb was that, if I didn’t like it, it was probably made by a human. That got me a five out of five score but it may just have been dumb luck. If it wasn’t, it suggests that human artists still remain better at challenging us, at stretching our tastes, than AI. But for how long?

In my view, however, the really interesting question, in du Sautoy’s talk, brings me back to my socks. Should we worry whether a creative work was made by AI or HI (human intelligence)? If we can’t tell the difference, does it matter? Does a computer-generated Bachian fugue or Shakespearean sonnet somehow make the originals less sublime and diminish us as human beings? Du Sautoy would claim not. Instead, he would claim that AI-creativity is a tool that will take us both (humans and computers) to new heights not reachable on our own.

Ironically, in putting forward these arguments, du Sautoy actually made a strong case that it does matter whether creations are of-human or of-machine. As he rightly emphasised, our creations allow us to glimpse the mind of another person—to partially see the world as they see it—and AI-creations may, similarly, allow us to glimpse the minds of our machines. Perhaps, our computers can even be taught to understand us better by looking at our own works of art. Art, as a way to recognize intelligence and as a way for intelligences to communicate is a fascinating idea at the core of du Sautoy’s book. I’m going to have to buy a copy and find out more.

Written, in vivo, by David Waltham. Photo © Fernando Bagué.

Marcus du Sautoy, author of The Creativity Code, appeared at Swindon Arts Centre as part of the Swindon Spring Festival, 15th May 2019.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: