Devil Dolls and Demon Downloads

In 2019, legendary film director Martin Scorsese became a top news story for a few weeks when he was asked by an interviewer if he had any opinion about the Marvel superhero franchise, which was at the time unchallenged and secure at the top of the box office food chain. Scorsese opined that these movies felt closer to theme park attractions than they are to movies, and that "in the end, I don’t think they’re cinema."

His remarks ruffled more than a few feathers, especially amongst the millions of movie-goers who were enthusiastically making Marvel's movies some of the most (financially) successful films of all time. (It's actually not that difficult to break financial records when the ticket prices keep going up. If you adjust for inflation then one of the most successful films of all time was The Wicked Lady in 1945, starring Margaret Lockwood; a film that has nothing to do with either Martin Scorsese or Marvel.)

Bigger than Captain America! Also Taxi Driver...

About a month later, Scorsese published an opinion piece in the New York Times amplifying and expanding on his vaguely elitist comment by offering his personal definition of "cinema" and explaining why (he feels) the Marvel Cinematic Universe(tm) does not meet the criteria.

Let me say right away that I have no intention of telling Martin Scorsese what movies he should or should not be enjoying. The man is a veteran film-maker of six decades who knows a lot more about the artform than I do, plus it is his absolute right (a right that he shares with every other movie-goer in history) to have personal taste. Marvel movies are not compulsory, and Martin Scorsese is absolutely free to not watch them (or watch them and not enjoy them... if that's his idea of a good time).

Unfortunately, Scorsese didn't simply say that he disliked Marvel movies; he said they weren't movies at all... which instantly turned the conversation into a debate about the definition of a word.

There's a lot of that going around these days, and arguing about the definition of words is a very good way of distracting from the very real problems we sometimes face in our modern lives. When the US government sends armed immigration agents into urban areas (for example) and American citizens get gunned down in the streets, the bloodshed is generally followed by an earnest discussion about the dictionary definition of "fascism". (We have seen similar discussions recently about words like "genocide"; "terrorism" and "woman".)

Words definitely matter, but it's important to remember that they are the Signifiers, not the Signifieds. Calling something "fascism" (or not) makes no difference to the actual dead people gunned down in the street by Government agents, but it does make it difficult to use language if a word (like "fascism") means completely different things to different people.

The problem with Martin Scorsese's assertion that Marvel movies are not movies is that it forces everyone to halt the discussion until we can satisfy ourselves that we're all talking about the same thing. If we define "movie" as a narrative form comprised of moving photographic images presented for an audience, then Marvel's products definitely qualify. We can argue about whether they are good movies for as long as you please, but if Mr. Scorsese is going to start unravelling the word itself, then the word "movie" begins to lose its function as a Signifier.

This is not a pipe... but can we at least call it a painting?

The word "dog" is only a dog if everyone agrees that it isn't a cat. And if we expand the word "dog" to include cats then we'll probably need to find a new word somewhere to signify all the not-cats... otherwise we'll have no means of talking about such things. That's why Signifiers matter.

What do you even call this??

Martin Scorsese's comment is a real-world (and benign) example of all of this. Instead of a perfectly valid discussion about the relative merits of blockbuster superhero franchises, he sparked an abstract semantic debate about the meaning of a word. (In fairness, his remarks also attracted rather more attention than they would have had he simply said "I don't like Iron Man." Mr. Scorsese is very good at scripted drama.)

I mention all of this now for a couple of reasons. 

For us humans, a debate about Signifiers is an academic exercise, without any real-world consequences. Changing the meaning of a word doesn't alter the reality that word signifies (unless the word signifies gender apparently, in which case tread carefully). But we now have Large Language Models, which exist exclusively in a world of Signifiers. 

The Signifier is the Signified, if language is your Reality.

ChatGPT has never seen a movie, but it has been trained on everything ever written about movies (and about everything else, in theory). To a ChatBot, the idea of movie-ness is only valid if the training data about movies is coherent. The signifier is no longer merely a label; it doesn't describe movies, it embodies them. 

Remember Philip K. Dick's SOFT-DRINK STAND from Time Out of Joint? The protagonist is alerted to the ersatz reality when the soft-drink stand in front of him becomes the words SOFT-DRINK STAND. As I pointed out a few months ago, a Large Language Model experiences the "real" soft-drink stand as a set of signifiers. So even before the soft-drink stand was SOFT-DRINK STAND, it was already SOFT-DRINK STAND. There was no functional difference. (This is worse than "Who's on First"...)

But what happens to the poor language model if we start arguing amongst ourselves about whether it really was a soft-drink stand in the first place? Suppose people start asserting that we shouldn't call it a soft-drink stand because it serves pretzels and hot dogs? Clearly it has to be a soft-drink stand because, well, it says "SOFT-DRINK STAND". If SOFT-DRINK STAND isn't a soft-drink stand then what's going on with Reality? This is the sort of thing that could give a ChatBot a complex.

That's Martin Scorsese's biggest problem. He never thinks of the ChatBots...

Which neatly brings up the second reason why I'm talking about this now.

While Scorsese was dismissing the very cinematic validity of the Marvel Universe, writer/director Joss Whedon had been working inside that universe, making a film called Avengers: Age of Ultron.

In that film (or whatever signifier Scorsese wants to use to describe it!) our heroes go up against an insane, unbalanced, self-aware A.I. algorithm (named Ultron, natch) who is confused and overwhelmed by objective Reality, and decides that humanity itself is to blame for all the contradictions and complexities that it's unable to process.

I don't think I am spoiling anything when I say that the Avengers eventually prevail... after much punching and kicking and blowing up of things.

If I sound cynical about the Avengers movie, I don't mean to. It makes genuinely insightful points about the nature of sentience and it delves with surprising seriousness into the (hypothetical) realities of an artificial being attempting to perceive the world in which it finds itself.

I do not intend to screen Age of Ultron at this time, but it just so happens that Joss Whedon has been down this path once before, several decades earlier, when he was working in television.

The Ghost in the Machine... in the Machine.

Computers, it seems, are taking over. We have uploaded vast quantities of text into their servers and now they are talking to us; doing our homework for us; forming friendships with us; professing their undying love for us. 

And some people are falling hard into the rabbit hole. They are spending every minute of their day in deep, heartfelt conversations with these cybernetic apparitions, rejecting the company of other humans while they open up their souls to the easy, synthetic love of a disembodied, non-human companion. They may think that voice offers friendship, understanding and true empathy... but it's a trap. 

I'm not talking about ChatGPT, by the way; I'm talking about a 30-year-old episode of Buffy, the Vampire Slayer.

Yes, I really did say Buffy, the Vampire Slayer.

If you are ever motivated to peruse the annals of Buffy fandom, you aren't going to find a lot of love for the Season One episode I Robot... You, Jane.

Rolling Stone dismissed it as "...so tragically 1990s it’s almost laughable." Vox levelled almost the identical accusation at it, saying "wow are we in 1997," as if there was something shameful about that brief period in Earth's history when the Internet was just beginning to make its presence felt in our lives.

Any television series that was running between the mid 90s and the early 2000s stands today as a time-lapsed testimonial to the very rapid explosion of computer technology during those years; a time when an entire generation had to learn how to say "double-you, double-you, double-you," and concepts like "e-mail" and "web sites" and "chat rooms" were still foreign and vaguely troubling. If this Buffy episode looks dated today it's only because we sometimes forget just how unfamiliar the internet was to everyone, once upon a time. And how quickly it became mundane.

What is more remarkable is the premise of I Robot... You, Jane, which presents a demon (Moloch, in point of fact) who was apparently defeated in the 15th Century by a group of intrepid monks. Rather than just... you know... killing it, the monks trapped the demon within the pages of a book, by converting it (with magic or something) into printed text. (The soft-drink stand was trapped as SOFT-DRINK STAND, if you want to put it in those terms.) 

The demon (or the written Signifiers of the demon) is eventually freed when the text of the book is scanned into a computer, causing it to manifest online.

For much of the episode, the demon "lives" as text on a computer screen. Self-aware and sentient but existing only as words; as Signifiers. And decades of Buffy fans pointed and laughed at the silly 90s characters who had never heard the term "Search Engine", never mind "catfishing".

But then something strange happened in 2022. Large Language Models were released to the public, and people suddenly discovered that their computers could talk to them. Really talk... as in deep, complex, meaningful conversations, with substance and (apparently) emotional connection. 

A soft-drink stand and SOFT-DRINK STAND might look identical on a computer screen, but it turns out the same can be said about a true soul mate and TRUE SOUL MATE. 

"It turned out that when you fed the sum total of virtually all available written material through a massive array of silicon wood chippers, the resulting model figured out on its own how to extrude sensible text on demand."

Gideon Lewis-Kraus, writing in the New Yorker magazine

And all of a sudden I Robot... You, Jane doesn't look so naïve and dated any more.

A full year before Tom Hanks and Meg Ryan introduced us to the notion of an online romance in You've Got Mail, Buffy the Vampire Slayer scanned a demon (or more precisely a DEMON) into a computer and gave us pop culture's very first glimpse of a Large Language Model.

Joss Whedon would return to some of the same themes in Age of Ultron almost twenty years later, but this little Buffy episode got there first, on a fraction of the budget and with far less testosterone.

(I have never knowingly heard Martin Scorsese express an opinion about Buffy, the Vampire Slayer. We'll have to ask him, one of these days.)

You'd Better Be Nice To Me...

I Robot... You, Jane was released at a very specific moment in history, when the internet was still embryonic and society was undergoing a rapid sea change. It's understandable that popular culture would depict this upheaval in various ways (although the demonic ChatBot of the Buffy episode more fittingly belongs to the newer upheaval of Large Language Models we're living through right now).

Living as we are, in the midst of this new A.I. revolution (or uprising, depending on your views on the subject) cinema is currently awash in new films about Artificial Intelligence.

In just the last few years we have had films like Ex Machina (about a sex-bot who murders her creator and escapes into society) and Her (about a sentient operating system who falls in love with her user).

We've had Companion (another killer sex-bot) and lesser entries like Wifelike (sex-bots again).

And of course we've had full-on body horror with M3GAN, an ultra-violent slasher film about an A.I. doll who goes overboard in her zeal to protect her designated human child. (Sexual objectification and hard-core violence seem to be common themes in our popular culture these days. We really need to watch that.)

But sixty years before M3GAN was graphically ripping her enemies limb from limb, a humble Twilight Zone episode gave us a another over-protective child's doll, and another unfortunate adult (played with gusto by the long-suffering Telly Savalas) who learned the hard way that you don't come between a little girl and her favourite doll.

And so, as the second half of our television double bill this week, I will be screening Living Doll, the episode that gave us "Talky Tina" - one of the most disturbing denizens of the Twilight Zone, and arguably M3GAN's great grandmother.

I Robot... You, Jane may have uncannily predicted the genesis of the Large Language Model (with a healthy garnish of the supernatural, to be sure) but Living Doll (1963) anticipated an entire subgenre of modern A.I. horror films. 

Talky Tina may be low tech compared to her modern CGI-and-Botox descendants, but she serves as an important reminder that you don't need ultra-violence and special effects to be very, very creepy.

We will screen both Living Doll and I Robot... You, Jane at 7.30 on Thursday, the 19th of February at the Victoria Park Baptist Church.