Columnists
An avalanche of drivel
Did you ever watch that 1996 Tom Hanks movie That Thing You Do!? It was about a high school rock band that became an overnight sensation. It featured some catchy songs, which sounded for all the world like the work of the Beatles. But as you analyzed it, you realized it wasn’t. Someone must have spent hours and hours composing ersatz Beatles songs (and doing a darn fine job). Same scenario with the 1978 movie All You Need is Cash (The Rutles), brought to us by some of the Monty Python imaginations.
Those song efforts threaten to become historical curiosities. Nowadays, you can put a computer neural network to work learning what a Beatles song sounds like, and then get it to craft a Beatleslike song. And, theoretically, not just one of them—dozens, hundreds, millions, even trlllions, at the flick of a command switch. Let’s call it “AI music.”
Such, at least, is the promise held out by Google’s Project Magenta, which aims to use AI music to give artists a whole arrray of new tools to create art with. The Magenta project is by no means the first AI music out of the starting blocks; you can already find examples of jazz and classical pieces generated as AI music on the Internet, as well as ‘butter or margarine’ tests between machine and human created poetry. But the fact that Google is getting into the creative arts business in a serious way—and sharing its secrets with developers—means that progress will no doubt be made quickly.
Of course, some would question whether this is a Good Thing. Do we need trillions of new Beatles-like songs? Well, maybe that case is the exception. How about a virtually inexhaustible supply of foul-mouthed rap music: now you’re talking about a serious avalanche of drivel. How many different ways are there to rhyme with “pluck” and “switch”? (The 19th century philosopher John Stuart Mill complained that the number of ways music could be expressed was finite. Technically, he may have been right, but I’m relieved to note that the limit has not been reached in the following couple of centuries—with the notable exception of contemporary country and western music.)
But replicating existing creations is just the beginning. Once the computer network has taught itself to recognize both a Beatles tune and rap music, there is no stopping it. It can be asked to create music that is a hybrid of the two. It can be asked to create an instrument that sounds like both a banjo and a set of bagpipes. Or something still worse. The opportunities for an AI music Dr. Strangelove are immense.
Some can’t wait for the onslaught of AI music. Says David Cope, a retired professor at the University of California- Santa Cruz and pioneer in computer generated music, “The computer is just a really really high class shovel. I love this new stuff and want it to come fast enough so I’m not dead when it happens.” Commentators say that film music—music used to enhance visual effects—will serve as a good testing ground for the development of AI music. And you can see a respectable argument that human music creators don’t have to lay down and die just because AI music has arrived; it’s but another creative tool at their disposal.
I prefer to see a few storm clouds darkening the otherwise sunny horizon. If the computer can generate trillions of Beatles-like songs in an instant, where does that leave the starving composer? How does he justify his devotion to his craft? Why bother struggling to create when a computer can come up with substantially the same outcome almost effortlessly? Will philistines now have the arguments they need to tell creative types once and for all to ‘grow up and get a job’?
Perhaps jobs will open up in sorting through the avalanche of drivel to find the few jewels that are no doubt in there somewhere. Perhaps it will be necessary to develop an accreditation program; so that just as you can get your honey ‘certified organic’ by an independent agency, you will be able to get can get your music ‘certified AI free.’ Or, like confectionery and peanuts, it might be necessary to add warnings that compositions ‘may have come into contact’ with AI. And just think of the copyright complications. Is a new Beatles-like song to be credited to Lennon & McCartney, to ‘AlphaBeta- Magenta Network 3.7’, to its human operators, or to some messy combination of all of the above? I see lawyers getting richer off this one.
Maybe the hope for civilization lies in the very fact that an avalanche of drivel will arrive. People will more quickly tire of rap music and contemporary country and western music and they will die a richly deserved, speedier death. The Beatles, of course, will live on regardless.
Comments (0)