September 2023

Arts & Letters

Workers’ singularity: AI and the future of art and labour

By Shane Danielsen

Writers Guild of America protest in New York, May 10, 2023. © Richard B. Levine / Alamy

The Hollywood writers’ strike has put a spotlight on the impact artificial intelligence may have on artistic endeavour

The Writers Guild of America strike, which began in May and may well continue into next year, was originally a response to economic hardship. Specifically, to the cascading inequities of the US television industry, which, despite offering ever more plentiful outlets for screenwriters (599 English-language scripted shows aired in 2022, according to FX Research, up 7 per cent from 2021), has managed to turn what had been a comfortable middle-class profession into a red-in-tooth-and-claw manifestation of the modern gig economy – a frantic, patchwork hustle of shorter contracts, fewer positions and reduced wages, with even established writers scrambling to earn enough each year to qualify for healthcare. 

About a month before the strike began, however, the discussion abruptly pivoted. Yes, all those things were important, but there was also another, even more significant issue to address. The growing sophistication of artificial intelligence modules had taken many by surprise – including those responsible for inventing them. (“I thought [machine intelligence] would happen eventually, but we had plenty of time: 30 to 50 years,” said Geoffrey Hinton, the British scientist whose work developing neural networks paved the way for platforms such as ChatGPT. “I don’t think that anymore.”) Suddenly social media was awash with examples of AI-generated storytelling, some of it even accompanied by rudimentary visuals. 

None of these miniature opuses could be called good. Many were comically inept; a few were frankly nightmarish. Nevertheless, the implication was clear: these were just the fledgling steps of an emergent technology, one not only growing more adept by the day, but with the clear potential to supplant writers and eventually replace them altogether. Perceptibly unnerved, the WGA demanded to know the Alliance of Motion Picture and Television Producers’ intentions regarding AI. Was it something they should be worried about?

To this question, the Alliance responded with blithe equanimity. Yes, they purred, of course they were exploring artificial intelligence. In fact, many studios and streamers, platforms such as Netflix and Amazon, had already quietly established in-house AI divisions. Not to replace writers – good heavens, no – merely to augment and assist them. And what’s more, these companies had no intention whatsoever of reversing this decision.

With that statement, the threat of wholesale obsolescence became both concrete and imminent. And industrial action moved from being a negotiating tactic to a practical necessity.

Renegotiating their own contracts with the AMPTP, the Directors Guild of America soon capitulated, reaching a deal that made “significant gains across key economic and creative rights while reaffirming the critical role of DGA directors and their teams” – proving, yet again, that you can never underestimate the self-interest of directors. The Screen Actors Guild, however, saw the crisis for what it was. They looked at the state of the technology, considered the avarice of the studios and streamers, and realised that their position was uncomfortably exposed. That their members’ faces and voices – the marketable, singular essence of themselves – were easily digitised, and therefore infinitely replicable. Once this was achieved, the actual human being was superfluous. “This is a power grab, pure and simple”, one “well-established actress” told Deadline. “We see what’s coming. They can’t pretend we won’t be used digitally or become the source of new, cheap, AI-created content for the studios.” 

In a Twitter thread that went viral in May, actor and writer Justine Bateman put it even more bluntly: “AI-written scrips [sic] & digitally-scanned actors (image and/or voice). Both already exist. Some talent agencies are actively recruiting their clients to be scanned. You choose the projects and get 75 cents on the dollar.” Depressing as that sounds, it’s a rather better deal than that offered extras (or “background artists”, to use the modern parlance), who under an AMPTP proposal would have their likeness scanned to be used in any way, on any project, without consultation or consent, for all time. In return, the performer would be paid for a single day’s work – which is to say, the princely sum of US$219.

Suddenly the battle for better conditions had transformed into something else: a struggle to survive, a last-ditch stand against a vast, existential threat. (“I believe this is the last time any labor action will be effective in our business,” Bateman concluded. “If we don’t make strong rules now, they simply won’t notice if we strike in three years, because at that point they won’t need us.”) And so Hollywood, for so long a mirror to the culture, found itself for once at the tip of the spear: the first, bitter skirmish in a war that will, over the next few years, impact almost every field of industry and human endeavour.

The encroachment of AI upon our lives is a subject almost too vast to contemplate; its broader possibilities, both positive and negative, are currently exercising thousands of lawyers, philosophers, economists and think tanks around the world. But given the steady pace of breakthroughs – the almost daily sound of expectations being exceeded – it seems reasonable to assume that we find ourselves on the verge of something momentous, a genuinely transformative moment in the history of our species. 

“It’s like the birth of the internet,” someone said to me recently. (Tellingly, it was the fifth AI-related conversation I’d had that day.) No, I replied, it’s deeper than that. It’s like the discovery of fire. In that there will be a Before This and an After This, and those two states will resemble each other obliquely, if at all.

Should recent developments continue apace, it’s reasonable to conclude that sometime within the next decade or so – perhaps in 10 years, perhaps in five – homo sapiens will cease to be the presiding intelligence on this planet. For the first time in the history of our species, something else, mentally at least, will be in charge. How that plays out is anybody’s guess. It could be Terminator or it could be Shangri-La, a dystopia or a paradise. But what’s not in dispute are the short-term economic consequences. The displacement of labour, as tasks are automated and workers retrenched, will be without precedent, and the effect potentially catastrophic. Architecture, town planning, graphic design, project management, diagnostic medicine, the work of solicitors and subeditors, technical writers and accountants and coders… all of these professions will be depopulated, emptied out in much the same way (to paraphrase Hemingway) that one goes bankrupt: first gradually, and then suddenly.

This process, the migration of jobs from human to machine, is already under way. In July, the Indian startup Dukaan, a large-enterprise eCommerce platform based in Bangalore, replaced 90 per cent of its support staff with an AI-powered chatbot called Lina. And while its chief executive Suumit Shah made the customary noises of regret – the decision, he wrote, was “tough [but] necessary” – he couldn’t resist boasting, in the very same tweet, that it had also reduced his firm’s operating costs by 85 per cent. Other companies are rushing to follow suit. A recent report from Goldman Sachs (“The Potentially Large Effects of Artificial Intelligence on Economic Growth”) forecasts the loss of some 300 million jobs over the next two decades.

Even the most sanguine outcome to all of this – a kind of post-scarcity utopia, in which a universal basic income is in place and groundbreaking discoveries seem mundane, and things such as cancer and Parkinson’s disease are just distant memories – will require, at the very least, a comprehensive redefinition of what we understand by terms such as work, leisure and remuneration. A wholesale renegotiation, in other words, of the social contract, if for no better reason than the fact that you can’t put millions of people out of work and expect society to continue functioning. (At the height of the Great Depression, less than 25 per cent of the American workforce was unemployed.) To put it in the most moderate terms, the global consumer base will collapse. Less politely, there will be heads on poles. 

Right now, we’re experiencing so-called generative AI: deep-learning “large language models”, capable of generating text, images and other content based on the vast array of data they have been trained on. They are essentially predictive mechanisms – sufficiently fluent in received wisdom to mimic what we understand as sentience. There are many parallel systems being developed around the world, both by state actors and private interests, in a kind of informational arms race. Each of them is busily “scraping” the internet for every piece of text ever published and every image ever uploaded, in order to expand their vocabulary and refine their responses. (Though this has recently backfired, as the LLMs’ hunger for information has led them to plunder AI-generated web content, resulting in a feedback loop that has seen ChatGPT’s responses grow perceptibly stupider: the tech equivalent, if you like, of mad cow disease.)

What comes next – the impending level-up – is AGI, or artificial general intelligence. Sébastien Bubeck, a senior principal research manager at Microsoft Research’s Machine Learning Foundations group, has defined this as “systems that demonstrate broad capabilities of intelligence, including reasoning, planning, and the ability to learn from experience, and with these capabilities at or above human-level”. And it may be closer than we think. Earlier this year, Bubeck and a number of his colleagues were given early access to OpenAI’s GPT-4, the latest iteration of ChatGPT’s multimodal LLM, and in a paper entitled “Sparks of Artificial General Intelligence: Early experiments with GPT-4”, they noted that “beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance.” GPT-4 has been publicly available since March 14. Google’s forthcoming Gemini system (expected later this year) is reportedly more powerful still.

The end result, the big prize, is ASI: artificial super-intelligence – a level of computational acumen that qualitatively and definitively surpasses human intelligence (“You need to imagine something more intelligent than us by the same difference that we’re more intelligent than a frog,” says Hinton), and which leads to a Singularity: a self-aware, self-perpetuating succession of smarter and smarter systems. This achievement is fraught with potential perils. Swedish philosopher Nick Bostrom’s famous “paperclip maximizer” thought experiment – in which an AI tasked solely with manufacturing paperclips casually eradicates humanity in order to more efficiently perform its duty – offers a frighteningly plausible hypothesis of what can happen when human and AI goals are unaligned. But really, we can’t predict what will happen if and when ASI is achieved, any more than… well, a frog could.

There are, of course, many questions attached to all this. But the one that has been troubling me most these past few months is both more personal and more immediate: faced with a future in which most popular art is largely made by machines, would most people even care?

In September 2013, Paul McCartney released “New”, the first single from a solo album of the same name. Produced by Mark Ronson, the song had a “jaunty, Beatles-esque stomp” (per the UK Daily Telegraph) and an equally retro arrangement: a fat Höfner bass line, discreet horns, a hint of harpsichord. But with one crucial difference: the voice singing it was that of a 71-year-old recording veteran, not a 22-year-old mop-top. A little hoarser, his upper range slightly circumscribed. Far from diminishing the track, this lent it a kind of sentimental power: there was an interesting tension, you felt, between songwriting and performance, innocence and experience.

In May 2023, a YouTube creator – “daelims09” – uploaded a new version of the song, in which they had used an artificial intelligence module to replace McCartney’s 2013 vocals with a simulation of 1964-era Paul. The integration was seamless; the result sounded utterly convincing. And – because they could – daelims09 also added something else: an AI-generated John Lennon to sing the song’s middle eight, some 42 years after his death. 

This is creepy and a little weird, but hardly unusual – YouTube is awash with similar creations, the cybernetic mashups of a new generation. (A cursory search will disclose AI Freddie Mercury singing “Thriller”, or an extremely convincing AI George Harrison doing Oasis’s “Don’t Look Back in Anger”, the latter’s backing track also altered to sound more period-appropriate – sparser guitars, drier drums.) Far more noteworthy, to my mind, was the fact that the comments beneath daelims09’s video (since removed, owing to a copyright violation) were almost entirely from diehard Beatles fans – not outraged, as I’d expected, but tremulous with gratitude. “This made me cry,” one wrote. “I think it helps that I think John & George both would’ve loved this song.” Another said: “This is unbelievable, just like a new Beatles song.” A third: “Thank you so much for making this!!!”

For anyone still wondering which way the cultural wind was blowing, this seemed to provide a pretty definitive answer.

The reality is, many people – perhaps even most people – don’t want art to challenge their expectations, or even to surprise them very much. They want what is familiar and comfortable. More of the same, more often. We can be arrogant about this, and say that it’s because the public are lazy, bovine and thick. Or we can be compassionate, and note that most people today are overworked, underpaid and understandably anxious about the future; they’re simply too exhausted, too worn down by daily life, to want to be challenged very much. But whatever your take, the result is the same: a preference for comfort at the expense of innovation and, ultimately, the eradication of traditional creators, made superfluous by the evolving dialogue between a consumer and their content. 

Given the astonishing refinement of image-generating AI software such as Midjourney and Stable Diffusion, it’s not unreasonable to expect that, within a decade, viewers will be able to order entire AI-generated movies on demand. Bespoke content, tailored to one’s demonstrated habits and preferences. You might want to watch a Charade-like romantic comedy, a breezy, stylish crime caper set in the Greek Islands circa, say, 1955, starring Margot Robbie and Sidney Poitier. Or a sci-fi blockbuster about rogue comet-miners trawling the Kuiper Belt, with Die Hard–era Bruce Willis and Humphrey Bogart… and maybe Dudley Moore as well, for some comic relief. Just enter your preferences and let the machine do the rest.

Will these films be excellent? Probably not. At their best, they will merely remind you, in a ghostly, fragmentary way, of other, better movies, previous things you watched and enjoyed – enough, at least, to algorithmically shape this one. But will they be good enough to pass the time, to fill a few lazy hours between dinner and bedtime? Absolutely. And that, for more people than we might want to believe, will be enough.

If this sounds far-fetched – or simply like too much effort – consider that entire generations of people raised on gaming have a much more interactive relationship with narrative than your average 60-year-old. And while the personally tailored movie is still a few years away, smaller refinements are already upon us. One example: while in Cannes in May, I saw a film called Vincent Must Die by the French filmmaker Stéphan Castang, a horror flick about a guy who wakes up one morning and finds that everyone – workmates, neighbours, total strangers – suddenly and inexplicably wants to kill him. Two days later, I read a story in The Hollywood Reporter that said the film had been acquired for US distribution by XYZ Films. The kicker, though, came in the following paragraph: “Flawless, a pioneering film technology company and leader in video translation, recently partnered with XYZ Films and Tea Shop Productions to acquire the rights to foreign-language films and translate them into English for distribution in relevant markets … Flawless implements TrueSync technology to create a perfect lip-synced visual translation of the acquired title.” 

Which is to say, the company will use an AI module to subtly change the faces of the actors onscreen, so that they can be seen to be “speaking” English (or Danish or Farsi or Wolof, depending on the territory). And to complete the illusion, the original cast’s dialogue will be sampled, in order to deliver these translations in their own voices. 

Goodbye dubbing artists, goodbye translators. Goodbye difference, in all its myriad wonder. A frictionless homogeneity beckons.

This is not to say that traditional modes of creation will disappear completely. There will still be people making art: writing novels, making movies and music and so on. Indeed, perhaps more of them than ever: we may have a lot of free time to occupy. But, barring a few exceptions, their audience will be minuscule, and the achievement devalued almost unto meaninglessness. (On August 7, Jane Friedman of The Hot Sheet, a newsletter dedicated to the publishing industry, reported that “As of today, there are about half a dozen books being sold on Amazon, with my name on them, that I did not write or publish. Some huckster generated them using AI.” The reputational damage, you suspect, is exceeded only by the general public’s indifference.)

Film and television are broadcast mediums, and, as such, have always been easily seduced by advances in technology. They may be lost, at least as we currently know them. Theatre, though, may not only survive but might actually prosper, provided it leans into its most handmade, artisanal aspects. Stories made by humans, for humans to enjoy, in a shared human space. 

That could work.

For now, the ruptures continue, mostly unnoticed. Indeed, just as I began typing the preceding paragraph, a friend sent me a link to a news story: “Zoom’s Updated Terms of Service Permit Training AI on User Content Without Opt-Out”. Minute by minute, click by click, we are perfecting either our factotums or our successors. One day soon we will know which.

Shane Danielsen

Shane Danielsen is a screenwriter and former artistic director of the Edinburgh International Film Festival.

There is nowhere quite like The Monthly. We are told that we live in a time of diminished attention spans; a time where the 24-hour-news-cycle has produced a collective desire for hot takes and brief summaries of the news and ideas that effect us. But we don’t believe it. The need for considered, reflective, long-form journalism has never been greater, and for almost 20 years, that’s what The Monthly has offered, from some of our finest writers.

That kind of quality writing costs money, and requires the support of our readers. Your subscription to The Monthly allows us to be the home for the best, most considered, most substantial perspectives on the state of the world. It’s Australia’s only current affairs magazine, an indispensable home for cultural commentary, criticism and reviews, and home to personal and reflective essays that celebrate and elevate our humanity.

The Monthly doesn’t just comment on our culture, our society and our politics: it shapes it. And your subscription makes you part of that.

Select your digital subscription

Month selector

From the front page

Kim Williams seen through window with arms half-raised

The interesting Mr Williams

At a time when the ABC faces more pressure than ever before, is its new chair the right person for the job?

Illustration by Jeff Fisher

Letter from Dunkley

As a byelection draws the nation’s focus to the scrappy suburb of the author’s childhood, a visit reveals the damage wrought by the housing crisis

Exterior of the Department of Treasury, Canberra

Tax to grind

Tax reform should not be centred on what we want, but on who we want to be

Rehearsal for the ABC TV show ‘Cooking with Wine’, March 13, 1956

Whose ABC?

Amid questions of relevance and culture war hostilities, the ABC’s charter clearly makes the case for a government-funded national broadcaster

In This Issue

Voice to Parliament rally at Prince Alfred Park, Sydney, July 2023

Demographics and the Voice

What past referendums tell us about the Voice to Parliament’s chances, and the political risk of the Coalition’s opposition

Microscopic image of Multiple Myeloma cells

Science fiction in university labs?

The case of UNSW and an “anti-cancer superdrug” highlights issues with self-regulation in universities about what constitutes research misconduct

Illustration by Jeff Fisher

The woodsplitter

A gruesome accident becomes a family’s touchstone moment of love and resilience

Detail of the cover of the first issue of The Phantom, published September 1948

Purple reign: The 75th anniversary of ‘The Phantom’

The longevity of the world’s first costumed superhero reflects an Australian publishing success story

More in Arts & Letters

David Malouf, March 2015 in Sydney

An imagined life: David Malouf

Celebrating the literary great’s 90th birthday with a visit to his incongruous home of Surfers Paradise to discuss a life in letters

Tony McNamara in New York City, January 2024

Pure things: Tony McNamara

How the Australian screenwriter of ‘Poor Things’, who cut his teeth on shows such as ‘The Secret Life of Us’, earnt his second Oscar nomination

Jeffrey Wright in ‘American Fiction’

The dread of the author: ‘American Fiction’ and ‘Argylle’

Cord Jefferson’s satire about Black artists fighting white perceptions of their work runs out of ideas, while Matthew Vaughn’s spy movie parody has no ideas of its own

U2 performing in the Las Vegas Sphere

Where the feats have no name: ‘U2:UV’ at Sphere

It’s no surprise it took U2 to launch post-stadium rock via a spectacular immersive show within the technical marvel of Las Vegas’s newest venue

More in Film

Tony McNamara in New York City, January 2024

Pure things: Tony McNamara

How the Australian screenwriter of ‘Poor Things’, who cut his teeth on shows such as ‘The Secret Life of Us’, earnt his second Oscar nomination

Jeffrey Wright in ‘American Fiction’

The dread of the author: ‘American Fiction’ and ‘Argylle’

Cord Jefferson’s satire about Black artists fighting white perceptions of their work runs out of ideas, while Matthew Vaughn’s spy movie parody has no ideas of its own

Michael Fassbender in ’The Killer’, sitting in a room cross-legged on a mat, wearing black gloves

Into the streaming void: ‘The Killer’ and ‘They Cloned Tyrone’

David Fincher’s stylish pulp and Juel Taylor’s SF-adjacent satire are the latest riches to be taken for granted in the ever-ready, abundant world of Netflix

Nick Cave performing with The Birthday Party at The Venue, London, 1981

The candles flicker and dim: ‘Mutiny in Heaven: The Birthday Party’

Ian White’s documentary captures the incendiary trajectory of the seminal Melbourne band at the expense of the inertia that fuelled it

Online latest

Osamah Sami with members of his local mosque

In ‘House of Gods’, Sydney’s Muslim community gets to be complicated

Plus, Barnaby Joyce shines in ‘Nemesis’, Emma Seligman and Rachel Sennott deliver ‘Bottoms’, and Chloë Sevigny and Molly Ringwald step up for ‘Feud: Capote vs. The Swans’.

International Film Festival Rotterdam highlights

Films from Iran, Ukraine and Bundaberg were deserving winners at this year’s festival

Two women on a train smile and shake hands

‘Expats’ drills down on Hong Kong’s class divide

Plus, Netflix swallows Trent Dalton, Deborah Mailman remains in ‘Total Control’ and ‘Vanderpump Rules’ returns for another season

Image of a man playing music using electronics and the kora (West African harp)

Three overlooked albums of spiritual jazz from 2023

Recent releases by kora player John Haycock, trumpeter Matthew Halsall and 14-piece jazz ensemble Ancient Infinity Orchestra feel like a refuge from reality