Into Uncanny Valley

Pinocchio the Pop Star

[This is edited and expanded from an earlier piece]

Before I start this piece, I must explain the concept of the uncanny valley, a descriptor coined by a Japanese roboticist called Masahiro Mori as early as 1970. 不気味の谷, Bukimi no tani genshō, means 'valley of eeriness’ and has been transliterated as ‘uncanny valley’. Mori saw that people were more sympathetic towards robots if they appeared human and displayed human characteristics and behaviours. This worked up to a certain point until the sensation of the eerie took hold and then the humans began to feel disturbed, troubled, and sometimes freaked out, even scared. This is the Uncanny Valley. 

“I have noticed that, in climbing toward the goal of making robots appear human, our affinity for them increases until we come to a valley, which I call the uncanny valley.”
— Masahiro Mori

Mark Fisher in his book The Weird and The Eerie from 2016 provides a definition of the eerie: “The eerie […] is constituted by a failure of absence or by a failure of presence. The sensation of the eerie occurs either when there is something present where there should be nothing, or […]  there is nothing present when there should be something.”

Another aspect here is one of agency. According to The Stanford Encyclopedia of Philosophy, “an agent is a being with the capacity to act, and ‘agency’ denotes the exercise or manifestation of this capacity.” Essentially AI is supposed to be a tool, a sort of puppet, as Pinocchio was created from wood by the lonely Geppetto. Yet Pinocchio wanted to be a real boy in his stumbling steps toward some kind of agency.

Carlo Collodi’s novel was published in 1881 in Italian. The story follows a wooden puppet who longs to become human, a real boy. Lacking the natural core of a fleshy human Pinocchio embarks on a series of adventures where he learns morality and ethics, freewill and choices, the difference between right and wrong, and the consequence of these choices. It is unclear if he becomes human or remains a  simulacrum constructed from the bole of a tree.

"Pretty eyes, pretty eyes!"

E.T.A. Hoffmann, a German author of strange fiction, wrote a short story called The Sandman in 1817. Freud references Hoffmann’s story in his 1919 essay ‘The Uncanny’. The liminal space between dream and waking begins the tale as a figure called The Sandman haunts young Nathaniel in his uneasy sleep. Nathaniel’s fiancée Clara reassures him that ‘all the terrible things of which you speak occurred merely in your own mind, and had little to do with the actual external world’. He meets his teacher Spalanzani’s daughter, Olympia, and falls in love with her. She is, however, an automaton, a perfect replication of a human female. In this story identities conflate and mingle, the reader is baffled; is the automaton Olympia an ideal version of Clara? Somehow a better and improved model of a woman in the eyes of Nathaniel than the reality of Clara?

I’m queasily reminded of plastic sex dolls.

There is certainly an element in rapid human perception which can identify the non-human and which then makes us uneasy. Try this experiment; watch The Polar Express - a film which turns up every year at Christmas on the telly and was released in 2004, when these new techniques were in their infancy. The technical and aethetic qualities were revolutionary but the emotional response is something else. Watch it closely. Using a relatively recent tool called computer-generated imagery (CGI) and additional animation technology the film has recognisable humans morphed into a new kind of animation, the first feature film to be created using the technique of motion capture technology which allowed animators to create a world without having to draw each frame as in traditional animation. A created universe with the human subject at a remove.

The Polar Express is widely cited as an example of the uncanny valley affect.

The problem is that what we see on screen is human enough, but not quite human enough, which leads to the creepy feeling and an alienation which can be hard to identify as such.

Now there are literary AI systems such as ChatGPT which will write full texts based on a few instructions within minutes or even seconds.

If someone has an idea for an essay or a poem or whatever and inputs the basic commands the AI will output a finished article in the blink of an eye.

Believe me, everything I write, down to the tittle and the final full stop, the spelling (mostly, because of the intricacies of the English language), the syntax, is written by myself. The poems are my own despite the inevitable borrowings . AI cannot really replicate the style which uses interior rhymes, odd metres, line endings which don’t always rhyme, the grime of experience, in short, individual voice. It cannot replicate you or me. So many writers have spent so many years gaining the literacy and it can’t be faked.

The poems produced by AI are so shite that they are often very funny. The rhyme scheme is clonky wonky with every line ending in a flat lame rhyme.

Twenty years after The Polar Express.

I have recently been given a demonstration of certain apps powered by AI which, basically, write music for both the pure consumer and the studio musician using advanced algo tech available at low cost. My mate Dave is investigating this new phenomenon and, man, this is seriously crazy. If you don’t know about this ‘advance’ then, like me, your jaw will drop. 

Allow me to explain about Dave. Years ago we were in a band, we made electronic music, early programmed music controlled by an Atari 1040. It was intially played direct as raw input then manipulated and arranged. The Atari had 1MB of memory; the latest Apple computer model has 768GB. The current iPhone has up to 512GB upgraded storage.

I’m talking about the AI called Suno and another one called Udio. No doubt there are others. What Suno does is a new kind of freaky magic. The user asks Suno to produce a full music tune using basic commands as to style, format, lyrics, everything. You don’t have to do anything; just send the commands and within seconds it presents a full track, arranged and ready to go. An option is to load a loop, however many bars, and Suno will output a finished track. To be fair, I can see human musicians getting some inspiration out of this if they’re stuck. It clearly uses thousands or even millions of tracks stored as templates. The output result: a tune in whatever style you choose. You want 1990s uplifting trance house w/ added vocals? There it is. You want 1930s close harmony? You got it. Want a cello or a string quartet? There it is.

Back to Dave. He uses modes and non-Western scales and odd tunings. Suno can’t deal with these and the process comes out seriously wonky. We drank cups of tea and smoked endless fags whilst discussing this and came to the conclusion that the generic pop chart sound as self-composed an studio engineered is in trouble. Suno will do it for you, sort of like cutting out the middle man. But original artists will thrive, as humans have always done, by being freaky original.

This is a gamechanger for production-line pop music, though. Machines can now make this. Is generic chart pop redundant in the face of this mad advance?

Still, cold AI will never replace the warm human spirit.

Musicians will have to up the game, establish new paradigms, become unique again and resistant to the copy machine.

There are a number of reviews of Udio and Suno online. Here are two: