News

Writers vs AI bots is more than a Hollywood drama

A confrontation broke out in Los Angeles this week as 11,500 writers for film and television went on strike. Screenwriters know all about confrontations: they are the second acts of three-act dramas, when the main characters face a crisis that only gets resolved at the end.

“You put them in the worst possible position they could ever possibly get into in their lives,” George Lucas, creator of the Star Wars films, once remarked of the second act. In this drama, writers have seen their pay erode in the streaming era, have lost the comfy conditions of broadcast television and fear their jobs will be taken by robots.

It is strange to find among a list of demands submitted to producers by the Writers Guild of America the call for only humans, not AI chatbots, to be allowed to “write or rewrite literary material”. Drawing attention to the fact that you think software could do some of your job is a bold gambit but screenwriters have vivid imaginations and reasons to be insecure.

The chances are slim of ChatGPT or another artificial intelligence agent advancing rapidly enough to write The White Lotus or Everything Everywhere All at Once in the immediate future. They sometimes make things up, which could come in handy, but it requires a lot of expertise to create dramas that engage viewers, structure them into episodes and polish dialogue.

Nor is screenwriting high among the professions that economists expect to be disrupted by AI soon: entertainment and media rank only in the middle of industries likely to be affected, with administration and legal services at the top, according to Goldman Sachs. Robots are not yet beating down the doors of the writers’ rooms on which Hollywood relies.

But the writers’ strike reflects a wider business phenomenon: the tilting of high rewards towards a few individuals at the top end, and the disruption of traditional paths to promotion with growing responsibility for apprentices. Machines are learning more than before, while humans learn less: it is not fanciful to fear a collision between the trends in future.

There was a moment during the last writers’ strike over pay in 2008 when studios let slip what was to come. Jeff Zucker, then chief executive of NBCUniversal, talked to the Financial Times of the “vestiges of an era that’s gone by and won’t return”. He cited the broadcast tradition of ordering many pilot episodes, turning only some into series and gradually winnowing them down.

It was an expensive habit but it provided steady work for writers on series that would run for 20 episodes or more, as well as residual payments for repeat showings on cable networks. Writers were employed for much of the year, and they worked not only on drafting initial scripts but rewriting during production and learning how dramas worked in the studio.

“Part of the job is to train the writers under you so they eventually understand what you do. How can you make a show if you’ve never been taught?” says Blake Masters, a writer and producer who created the drama Brotherhood. What felt to employers like financial extravagance meant consistent employment and long-term training and opportunity for the writers.

This diminished with the rise of streaming platforms such as Netflix and Amazon Prime, which preferred shorter series of six to eight episodes to give their subscribers constant novelty. They also paid highly to attract the top showrunners such as Netflix’s reported $150mn deal in 2017 with Shonda Rhimes, writer and producer of series including Bridgerton.

They moved to hiring writers on short-term contracts to develop new dramas in what are known as “mini-rooms”; these are disbanded before the show is commissioned. This approach not only pays writers less and limits residuals, but means they do not gain practical experience on productions; they must find themselves another mini-room project instead.

So writers cannot be blamed for fearing that studios will also exploit AI. Rhimes is not going to be replaced by a robot, but AI could be deployed in subtler ways. It is easy to imagine a future showrunner creating a story outline, getting an AI model that has been trained on thousands of scripts to rough out potential scenes, and finally giving the result to humans to polish.

The guild wants to stop its members’ work being used to train AI. It also wants them to be paid as much for rewriting AI output as if it were theirs. Both safeguards appear fair to me: without them, studios that fragmented the work of junior writers seem very likely to use AI to do some of their work for free. Technology could unleash a vicious cycle of creative deskilling.

The strike thus has broad implications. As Erik Brynjolfsson of Stanford University has written, the crucial question for AI in the workplace is whether it gets used to augment or automate labour. The first would raise productivity (and make better drama); the second would concentrate wealth and power in fewer hands. I hope for a happy resolution but the danger is real.

john.gapper@ft.com

Articles You May Like

Home price growth is back at pre-pandemic levels. Here’s what that means for buyers and sellers
The weakest links in the global economy are on the mend
Labour to promise greater devolution of power in England
The perils of speaking up at work too often deter staff from voicing concerns
A look at trailblazing women in public finance