Synthetic Symphonies Part III: The End

(This post is the third in a series on music and artificial intelligence and follows on from my last post. If you're new to the mailing list, it may be worth checking out the other two on our website.)

John Searle‘s 'Chinese Room' thought experiment is one of my favourite ways into thinking about technology and consciousness.

Imagine a sealed room with a narrow slot in its door, just big enough to slide pieces of cardboard in and out.

A native Chinese speaker is able to write sentences in Mandarin characters on pieces of card, slide them through the door, and receive detailed, natural responses as though they are communicating with another native mandarin speaker.

However, inside the room sits an American person with absolutely no knowledge of Mandarin. They simply have a detailed instructional guidebook that allows them to match the sequences of Chinese characters they receive with ‘response’ sequences from the guide.

They copy their chosen response onto a fresh piece of card, slide it back out of the door and wait for the next sequence of characters to appear and for the conversation to continue.

What’s interesting about this it that it gives us an insight into the nature of artificial intelligence.

The person inside the Chinese Room is conducting a conversation in Mandarin good enough to satisfy the person outside that they are ‘speaking’ it to another native Mandarin speaker.

However, the person inside the Chinese Room has no knowledge of what they are saying. They have no awareness of the direction the conversation is taking. For all they know they could be declaring their undying love, discussing the causes of the Second World War, trying to argue for the death penalty or simply asking about the weather. They may not even know what language they are using. They are just retrieving symbols from a database of information.

They are, in other words, emulating what it is to be a computer.

They are a blind processor of information; they receive an input and follow a set of instructions to create a satisfactory output.

What strikes me as uncanny about this idea is that, for the first time in human history, we have machine 'brains' that are able to communicate realistically through dialogue, art and music without any inherent understanding of what it is they are doing.

MuseNet - the software we discussed last time,  which composes original music by absorbing the corpus of human melody and harmony and training itself to predict the most appropriate note that should follow in any given sequence of notes - is like a musical Chinese Room. It has no awareness that it is producing music. It has no concept of the emotional effects its music might produce on a listener. It has no intentions. It doesn’t feel, think or know anything, including what music even is.

It is simply receiving an input and performing an incalculably complex series of pre-programmed actions in order to produce an output.

And I think, at its heart, this might be where the brooding discomfort I feel when thinking about music composed by artificial intelligence may lie.

Oscar Wilde’s novel The Picture of Dorian Gray begins with a series of infamous epigrams about the nature of art. Like much of Wilde’s work they are a deliberately infuriating selection of tangled ideas designed to tease and provoke. One of them simply reads: ’To reveal art and obscure the artist is art’s aim’. Wilde - at least the version of himself he was playing with in the preface to Dorian Gray - was an advocate of aestheticism. He purported to believe in the power of art for art’s sake; that beauty exists separate to the experience or intentions of its creator.

Wilde would, by this logic, believe that a beautiful symphony written by a blind algorithm with no awareness of its own actions should have just as much cultural value as anything written by Beethoven or Mozart.

Though I appreciate the thinking, and I certainly don’t believe that beauty needs to be created or intentional (any breathtaking view, or the smell of rain on a summer’s day quickly puts pay to that idea) my intuition has always been that it shouldn’t be.

I think the best art acts as a sort of window into the human soul - it illuminates, rather than obscures, our conception of the artist and allows us an insight into a psychic landscape that feels at once strikingly different but simultaneously familiar. In some sense it teaches us things we didn’t know about ourselves and what it is - or, at least, what it could be - to be human.

I’m not saying that one needs a detailed biographical understanding of the artist in order to appreciate their art. I don’t think you even need to know who the artist is.

Van Gogh’s burning gold Sunflowers, for example, certainly take on added drama when one knows a little about the life of Van Gogh; his later descent into madness, his schizophrenia, the vivid, unrestrained emotional turbulence of his life. Without that knowledge though, I think we would still recognise in the painting the spectres of all of those human qualities. We would be unable to ignore the humanity in every brushstroke.

We would see a simple object translated into paint in such a way as to tell us more about the painter of the sunflowers than the sunflowers themselves.

We can’t learn anything about what it is to be MuseNet from listening to a symphony composed by MuseNet. Mostly because it isn’t ‘like’ anything to be MuseNet; the same way that it isn’t ‘like’ anything to be a rock, or a plastic straw, or the coffee in my cup.

But I can’t help but be haunted by a darkly prophetic vision of the year 2030, or 2035, or 2045, or whenever it might be.

A day where the intra-neural micro-chip in my brain rouses me gently from sleep at the perfect moment between post-slumber and pre-waking.

My first tentative movements are read and received by my Smart-Mattress and downstairs, as if by magic, a filter coffee machine switches itself on.

And I hear it rising like the dreamy chatter of the birds: the greatest pop song ever written.

It’s a transcendental monument of words and music aligned in perfectly imperfect near-symmetry. It tells me all the things I never knew I felt about love and I know, right then, that I’ll be listening to it for the rest of my life.

Until I find out it was written and performed by no-one. It was simply processed and spat out by an unconscious machine who, by virtue of unparalleled processing power, has come to be able to know us better than we’ll ever be able to know each other. And, at the same time, is incapable of knowing anything at all.

And what do I think then?

Keep dreaming,

Rob

‍