While there are still so many unknowns around AI, we’ll only learn what audiences want by testing the water, says John Elmes

As a writer, I’ve been encouraged to believe that generative artificial intelligence is out to, at best, steal my work and, at worst, steal my job. It is a technology which has been met with utmost scepticism by TV creatives and was one of the central concerns of the 2023 WGA-led talent strike, which enshrined safety measures for writers. Today, creative unions are seeking compensation from companies who have scraped existing works – including journalism and scripts – to help train their generative AI tools.

John Elmes

John Elmes

Producers have, understandably, voiced concerns to me about the march of AI, notably the fact that Google has been scraping videos from YouTube to help train its Gemini model and Veo 3 video and audio generator. Their fears add up to the mind-blowing thought that, given buyers are increasingly looking at YT for new talent and fresh content ideas, if Google can potentially hoover up every video on the site to train its AI it could feasibly have access to every potential future format or programming concept.

I agree with all of their concerns, and it has fuelled my own fear of the tech.

AI is very much a part of the TV world now – but how our industry uses it is still being determined. Like most innovations, legislation is frequently much slower than the development of the tech itself, so lawful protections are coming, but still a little way down the line.

And the point is that like me, most people have had an introduction to AI based on fear, particularly of the unknown. Because that’s what quite a bit of the AI discussion revolves around: unknowns.

The inaugural SXSW London festival in June, had a lot of sessions on AI (by my unscientific count, 26 across four days) but it permeated even the TV industry-focused discussions, such as that between Banijay boss Marco Bassetti and Jordan Schwarzenberger, co-founder of Arcade Media, the management company that represents YouTube giants Sidemen.

Schwarzenberger posited an interesting idea – that younger viewers (the YouTube and social media platforms’ main demographic) prefer content with “lo-fi” production values, which are more “authentic” than the “finessed to the extreme” AI-infused video content.

But just a week later, on Banff World Media Festival’s AI panel, Raja Khanna, executive chair of virtual and mixed reality games company Dark Slope, told assembled execs the creator economy is producing YouTube and TikTok content easily with AI – and crucially that their peers are lapping it up, so the industry needs to get to grips with the tech fast if they don’t want to be left behind.

So which is it – (young) audiences don’t want AI-infused content or they’re so desperate for it that producers best starting thinking about using it in their productions? We’re back to unknowns.

About a month before SXSW, I wrote an article about a history doc looking at Viking emperor Knut, made by French indie ZED, which makes extensive use of generative AI to recreate historical sequences. The four-part series, which has been pre-bought by Sky History, uses the tech to – get this – “achieve a striking level of realism”.

I must admit, I thought parts of the industry would be up in arms about technology being used instead of employing real actors for reenactments – thereby undermining the docudrama sub-genre that has become incredibly popular in recent years – or at least that there’d be some scoffing about the irony of using technology to create “realism”. Not a peep. Something I was so sure of, turned out to be completely wrong – another unknown.

For me, all these things have sparked a slight mentality shift about AI. Yes, exercise caution with checks and balances, but maybe we should shed some fear. Let’s test the water and see what the industry – from practitioners to audiences – decides is the acceptable threshold for its use.

Only when there’s a massive outcry at someone pushing the envelope too far (whatever that looks like), will we start to see lines drawn. As producer and AI expert Benjamin Field said on that same Banff panel: “The audience will dictate.” Maybe we should let them choose?

  • John Elmes is deputy news editor and international editor