We have lately seen the world of creative types in quite a lather over artificial intelligence. The creatives think that big corporations are planning to replace them with AI writers, musicians, actors, and everything, and there will go the careers for the creatives. This was a significant issue in the recent writers’ strike.

Of course, the creative types are absolutely right. The big entertainment conglomerates are certainly looking at artificial intelligence and thinking how much money it could save them on writers, musicians, actors, and so forth. If we can just tell the bot to make a movie, what do we need with creatives?

That, at least is what the short-term thinkers in the entertainment business are thinking. If there are any long-term thinkers among the executives, which is unlikely, they are also considering which second career in manual labor they would like to adopt. If the ordinary viewer at home can just tell the AI bot, “Entertain me,” then what do we need with big entertainment conglomerates?

But what will entertainment by artificial intelligence be like? Dr. Boli was listening to some old French records the other day, and it suddenly occurred to him that we already know what AI entertainment will be like, because there is one branch of the entertainment business in which the robots have already taken over. Every popular singer is brought to us through pitch-correction software that adjusts the voice to what the robot thinks we want to hear, rather than what the singer actually sang. This is true even in so-called live performances.

Now, think for a moment what we would lose if we had had pitch correction decades ago. You can think of any singer you like, but Dr. Boli had these thoughts specifically when he was listening to Edith Piaf.

What made Edith Piaf such a legend? What made her stand apart from the hundreds of chanteuses who were popular in their day but have long since been forgotten? It would be very hard to answer that question, but Dr. Boli could easily give the wrongest possible answer. The wrongest possible answer would be something like this: “Edith Piaf was better than the rest because she was always exactly on pitch.” Surely Edith Piaf is at her most characteristic, her very Piafiest, precisely when she is deviating from the correct pitch. It is not that she is incapable of hitting the note right on: it is that she understands how to bend the pitches to her will. She was a legend because she could convey every shade of Gallic emotion, from cynical ecstasy to cynical melancholy.

You could think the same thoughts about any legendary singer of the past. What about Billie Holiday? Some day Dr. Boli will write an essay proving that it is heresy to say that there is such a thing as a “note” in jazz. Or what about Enrico Caruso? Surely the great Caruso was always on pitch! But listen to any Caruso record and you will hear that he was not on pitch: that the very thing that made Caruso stand out from the ordinary tenors of his time was his perfect instinct for twisting and bending the pitches away from the chromatic scale.

It is possible to use pitch-correction software to bend notes and deviate from the chromatic scale, but the very fact that we are discussing that possibility shows how many layers of artificiality have been stacked between us and the increasingly irrelevant human singer.

This is a preview of entertainment by artificial intelligence. We know what it will be like, because we have already seen what artificial music is like. Our AI entertainment will be perfect, in the same way that singing with pitch correction is ear-numbingly, mind-emptyingly perfect. It will be soulless, dull, and intolerably vapid. And the masses will love it, and they will want nothing else.


  1. tom says:

    It will be glorious someday when a few dozen human beings can sit in a darkened room and listen to a fellow anthropod sing what used to be called “a song.”

  2. RepubAnon says:

    “It will be soulless, dull, and intolerably vapid.”

    So, Muzak(r)?

  3. Occasional Correspondent says:

    Reminds me of Eubie Blake (ragtime pianist).  Other pianists would come to him with their own variations on the classics and standards of ragtime which he betimes recognized as arising from their inability to execute the piece as written.  Blake told them, Play it the way it’s written.  You don’t know how to break the rules yet.  Learn the rules, THEN you’ll be ready to start learning when and how to break the rules.  You take an easy way out (“your” way), you’ll never even master the rules, much less get beyond them.  So rote performance betters eased performance but rules-bound performance tops out below master performance.  (Even so, I’m sure Blake would agree with the sentiment “Please don’t shoot the piano player, he’s doing the best he can.”)

    In medicine there’s something that’s somewhat parallel.  Some surgical procedures are fiendishly difficult to perform but others are notoriously simple — appendectomy is often held up as a see-one-do-one-teach-one procedure (although others say it’s a four-beat sequence, see-one-screw-one-do-one-teach-one).  However, there’s more to medicine than manual proficiency.  Where a real doctor earns her pay is knowing WHEN to perform an appendectomy (etc.), not just how to do one, and it takes a lot more to learn that than to learn the handwork.

  4. Charles Louis de Secondchat, Baron de la Breed et de Montemiaou says:

    I find I must differ with the good Dr. Boli here, for which I hope he will pardon my little peculiarities. He seems undoubtedly right about the vapidity of mass-produced AI music (just look at the saccharine gunk produced by humans using the Midjourney AI) but I think he is probably wrong about AI finding Edith Piaf — or any human — an insurmountable obstacle.

    AI we know today is not at all like the AI Sci Fi writers told us about or even like the “AI” programs of early decades. Back then, we had every reason to expect that AI would just be a very complex precisely defined computer program — If (a){ do b;} else { do c}, etcetera. From this, many people very naturally concluded that AI would be the ultimate logician — a perfect reasoner in the same way that a calculator is a “perfect” mathematician. In such a world, it is very reasonable to assume that an AI will not be able to replicate Edith Piath, because it will only be able to sing perfectly, and Edith’s talent is knowing precisely when singing “perfectly” is for the pigeons.

    This is, as it turns out, not how machine learning works. Rather than explicitly coding everything, it is much more effective and easier to simply plug a bunch of examples into a large matrix of numbers and then put it through many rigorous rounds of artificial selection until we get the outputs we want.

    This method is NOT explicitly programmed at all — in fact, we usually don’t know how the final result works — but work it does and crucially, it often works in surprisingly human ways, because “human” is what it was trained on. AI writers will mimic human grammatical errors, AI movie makers will mimics static in old films, and AI musicians trained on the music of Edith Piaf will mimic Edith’s tendency to deviate from the correct pitch. You would not even *have* to tell the program to “deviate from the chromatic scale” — it would just do it because Piaf did!

    I think humans have a bad tendency to assume that there must remain *some* area of human expertise that will remain free of the unceasing encroachment of machine intelligence. This frequently express itself in the belief there is something unique about artistic expression that will remain forever out of the reach of the allegedly strictly-logical machines. Just think of the movie iRobot, where Will Smith asks rhetorically ” Can a robot write a symphony? Can a robot take a blank canvas and turn it into a masterpiece?*” Of course, now we know that far from being the last areas of human expertise the machines could not conquer, this was in fact one of their earliest conquests!

    The answer to the question “Can a robot sing as well as Edith Piaf?” is “Not yet, but they can already mimic the style and they almost certainly will be able to, given a few updates.” (Whether humans choose to make them sing like Edith Piaf is another matter.)

    If you do not want AIs that are superhuman at all possible human tasks, then I recommend not betting on some barrier the artificial mind cannot summit, but actively telling programmers ” I would actually rather not live in a world with superintelligient AI capable of outpacing all human effort, thank you, and if you insist on building it anyway we have a cozy jail cell with your name on it.”

    *(Amusingly, while looking this up, Bing’s AI agent helpfully informs me that the answer to “Can a robot write a Symphony” is Yes, and that the AIs having been doing this since 2011.)

Leave a Reply

Your email address will not be published. Required fields are marked *