WeeklyWorker

20.07.2023
Solidarity with writers and actors

Tinseltown’s hot summer

Writers and screen actors need to win in their strike against the media giants, says Paul Demarty. But our aim must be a better, a higher culture

It is perhaps counterintuitive that, despite being an industry built on the near-religious veneration of particular individuals, Hollywood is one of the most unionised places in the United States.

The unions are divided, regrettably, by craft; nonetheless, their coverage is extensive. You have the writers’, directors’ and (more dubiously) producers’ guilds (WGA, DGA and PGA); the Screen Actors’ Guild-American Federation of Television and Radio Artists (SAG-AFTRA); and, for the unsung heroes of Tinseltown - ‘below the line’, as they say - the International Alliance of Theatrical Stage Employees (IATSE) and the Teamsters. There is some unevenness in IATSE’s density, admittedly, but any one of these unions can bring the whole place to a halt.

It falls to the WGA and now SAG-AFTRA, this time, to remind us of this fact. The writers have been on strike for over two months now; last Friday, the actors guild joined them. The stars of the long-trailed Christopher Nolan blockbuster, Oppenheimer, walked out of the London premiere. Both unions are in sectoral disputes with the Association of Motion Picture and Television Producers (AMPTP), the industry body that periodically agrees an industry-wide contract. The points of dispute are very similar.

Both the actors and writers want decent residuals (basically, royalties) from streaming services, having fought tooth and nail to get them out of the Hollywood studios and over-air broadcasters in battles past. The AMPTP, on the other hand, sees streaming as an excellent opportunity to wind the clock back and keep more of the pie themselves. There are particular ‘workplace’ demands: SAG-AFTRA wants restrictions on self-taped auditions (the industry has been making ever more absurd and expensive demands on actors merely to get on the pile1), and the WGA wants to get its members access to sets (common in the studios and TV networks, but rare at the streamers, and important for honing craft and getting on in the industry).

What has dominated the headlines, of course, is the question of generative artificial intelligence. The WGA wanted an assurance that writers would not be replaced with AI models; the AMPTP refused, offering only an annual “consultation” on “disruptive technologies” - a clear signal that these desiccated philistines were just desperate to try exactly that. In relation to the use of digital likenesses of actors, the ‘compromise’ offered by the producers was, in the words of SAG-AFTRA negotiator Duncan Crabtree-Ireland, “that our background performers should be able to be scanned, get one day’s pay, and their companies should own that scan, their image, their likeness and should be able to use it for the rest of eternity on any project they want”.2

The industry lobby denies this, claiming that such usage would be restricted to “the motion picture for which the background actor is employed”. Supposing this is true, however, it is hardly reassuring. What counts as the “motion picture” here in a TV or streaming series context? An episode? A season? A whole series?

The matter of AI deserves closer consideration, if only because it appears here as a concentrated expression of wider social anxieties. When SAG-AFTRA president Fran Drescher told her members, “If we don’t stand tall right now, we are all going to be in trouble - we are all going to be in jeopardy of being replaced by machines”, she said no more than I have heard from worried software engineers, advertising creatives, photographers and many others in jobs that, while in some cases were quite precarious, were nonetheless largely insulated from the threat of obsolescence through technological advancement.

ChatGPT

We first need to be more precise about what we are calling AI - one of the most promiscuously overdetermined expressions in the English language. At issue is the latest generation of deep learning models. A model is essentially an algorithm: a deterministic series of computations that transform a particular input into a particular output. Given, typically in the case of popular products like ChatGPT and DALL-E, an instruction in plain English, the model produces some more text, or an image, or a video clip, or whatever. What goes on in between need not detain us in detail, though the details are interesting in their own right. The important part is the ‘learning’: the model, first of all, includes a corpus of works fed to it and initially manually categorised. It may then be given feedback on the outputs it produces: is this a good response to the prompt, “Paint me Donald Trump’s banquet of hamburgers in the style of da Vinci’s Last supper”? And, depending on if the operator gives a thumbs-up or thumbs-down, the algorithm will modify itself.

The salient facts here are, firstly, that human input is still required, but of a rather more mechanised and really subsumed sort than even the most exploited of junior screenwriters. The second is that there are two inputs - the prompt and the corpus of training data. The cardinal rule of all software systems - ‘Garbage in, garbage out’ - is pertinent. Applied to software engineering itself, some kinds of results from the currently modish, large language models are plainly useful. There are only so many ways to sort a list of numbers in ascending order, and far fewer good ones. ChatGPT ought to save us the effort. A great deal of the trade, of course, is not amenable to such automation, since it involves decision-making under conditions of uncertainty (when the customer says they want X, do they really want X, or Y?), non-obvious trade-offs between different goods (should the code be as efficient as possible, or as easy to understand as possible?) and so on.

Superficially, a creative pursuit like film-making resembles the latter part of the programmer’s craft more than the former. Yet, even in the best case, creativity is as much about borrowing as invention. Pablo Picasso is supposed to have quipped, “Good artists borrow, great artists steal”. From high culture (James Joyce’s Ulysses, for obvious reasons) to low (sampling in hip-hop music, say), new works grow in the composted remains of the old. Nor is this a distinctively modern phenomenon - Renaissance painters would train for years copying the works of masters, to say nothing of eastern icon painters, the orders of classical architecture, and so forth.

There is a distinctively capitalist mode of such ‘composting’, however. Capitalist firms use intellectual property to assemble their own corpus of works, from which they may produce revenue. They do this by monopoly of the means of production and distribution of cultural products, in which situation they may use the conventional methods of marketing to turn their properties into enormous branding empires. The characteristic cultural form of capitalism is not film, or music of any genre, or even television: it is advertising.

We need only think of Disney here: it has its well-known cartoon characters (Mickey Mouse, Donald Duck and friends); its vast back catalogue of cheerful animated classics; science fiction mega-franchises like Star wars; and the all-conquering cinematic behemoth of our day, Marvel Studios. All of these verticals have long since turned to eating themselves. The last Star wars trilogy was something like a billion-dollar equivalent of Gus van Sant’s shot-for-shot remake of Psycho, so obviously in hock was it to the original films. The animated classics are all being remade in what may loosely be termed ‘live action’ (in reality, green-screened to a plasticky sheen). Marvel’s latest big idea is the ‘multiverse’, which is a licence for infinite reboots, and perhaps finally the disproof of Alan Moore’s caustic description of such comic-book-film ‘universes’ as “sprawling, meaningless but at least still finite”.3 If good artists borrow and great artists steal, great cultural corporations steal from themselves.

In short, there are worse ways to think of Disney than as - alas! - an extremely analogue, extremely slow and extremely primitive AI. And it is this ‘machine’ that is the real threat (and its junior competitors, like Netflix and so forth). It ensures that the relevant AI systems will only be used for the worst, most scandalous ends: smothering creative endeavour, destroying the Hollywood craft unions, and drowning the rest of us in indistinguishable, aesthetically barren, but profitable ‘content’. For a glimpse of the future here, we need only take a look at the working conditions of the visual-effects artists, whose job it is to populate the green screens with actual action - a thoroughly globalised industry, it is poorly protected by the Hollywood unions.4

There is no reason to suppose that AI should not be a tool of writers or even actors (although, given the absurdly prissy guardrails that afflict the mass market tools,5 probably more specialised ones will be more useful here). In any case, there is nothing fundamentally more dystopian about the use of these tools than, say, using the I-Ching or Oblique Strategies and other generative methods, as Philip K Dick, John Cage and many others have profitably done in the past. The important matter - as it is in all labour disputes, at the end of the day, even those bizarrely involving millionaire celebrity ‘workers’ alongside starving extras - is who is in charge. The captains of this industry have film in a death-grip, and we wish them a humiliating defeat.

Victory?

The question therefore remains: in this pair of disputes, who is winning?

In the actors’ and writers’ favour, between the two - so long as discipline is maintained - production in the conventional centres of film and TV is pretty well shut down. Whether or not AI can replace actors or writers in the production of a certain sort of mechanically extruded cultural slop, the replacement has not yet happened. Even the best AI-generated scripts require a lot of punching up, even to make sense. An AI-generated model may look good in a crowd, but not in a tearful close-up. (Even video games are extensively reliant on motion capture to have any kind of immersive effect.) It was no surprise to see significant dips in the share prices of major media corporations on July 14 (Disney’s stock ticker almost exactly tracks the state of negotiations with SAG-AFTRA after the last month).

That said, it is the nature of the industry that there are long lead times. Strikes now will cause a dearth of content in two years. Corporations optimised for churning out cheap, predictable, thoughtless crap have ways to cushion these sorts of shortages. And, in any case, as we have seen, the industry is increasingly happy with endless exploitation of its existing products.

Given the clear intention on the part of the industry to break the power of these two unions, they have every interest in playing for keeps. Yet it is worth issuing a challenge of a sort: onward to victory, yes, but victory for what? To be blunt, it seems perfectly plausible to me that the assembly of moronic Marvel movies and worthless Netflix ‘content’ really is better left to computers, once they get the hang of it. (It is precisely this that marks it out as moronic and worthless.)

The victory of artists over misnamed ‘producers’ impliedly points to a whole different political economy of the arts as a whole. But that in turn points beyond the limits of ‘pure’ trade unionism, never mind the peculiar craft unions working above the line in Hollywood - to a true mass, working class movement, with its own embryonic political economy and cultural apparatuses, which can fight for artistic freedom from the rule of shareholders and accountants; for an end to intellectual property, which does basically nothing to help artists and creatives, though they are often deceived on this point; and ultimately for far wider involvement in the arts, broadly conceived, than capitalism permits.

  1. jacobin.com/2023/06/actors-sag-aftra-auditions-are-work-pay-labor-union-contract.↩︎

  2. www.theverge.com/2023/7/13/23794224/sag-aftra-actors-strike-ai-image-rights.↩︎

  3. slovobooks.wordpress.com/2014/01/09/last-alan-moore-interview.↩︎

  4. gizmodo.com/disney-marvel-movies-vfx-industry-nightmare-1849385834.↩︎

  5. Author Kat Rosenfield’s attempt to brainstorm a fictional murder with ChatGPT is an amusing case in point: twitter.com/katrosenfield/status/1672969824656322561. (DALL-E 2 also refuses to render my Donald Trump scenario above, which is a pity.)↩︎