AI: An existential threat?

The cultural sector has often had an uneasy relationship with new technology. While it undoubtedly brings new creative possibilities, it can also pose a threat to creative practitioners and their hard-won skills – especially in the case of generative AI, which is rolling out at a far greater pace and scale than any previous technology.

My thesis is that AI is misnamed, it may be artificial but it’s questionable whether it’s intelligent. So cultural practitioners should get stuck in, embrace the new technology and help shape it.

The revolution will be bumpy

We can learn lessons about how generative AI might impact the sector from the impact of emerging technologies on the creation and consumption of music over the past few decades. I’ll skip the introduction of recorded music and FM radio and go straight to the early 80s. Music lovers of a certain age will remember the apparent existential threat posed by the humble cassette to the music industry…

‘Home-taping is killing music’ was an aggressive slogan stamped on the inside of record sleeves. It didn’t. If teenagers hadn’t taped tracks off the Peel show, we wouldn’t have gone on to buy the LPs, go to the gigs and support a wave of creativity that’s changed the UK for the better – thanks to the humble TDK C90.

Two decades later the threat became more existential with the emergence of file-sharing and then streaming. But neither of these technologies stopped people making music. While the tech revolution caused a rocky period for the suits in the industry, they bounced back and are once again hoovering up the cash. Most musicians less so.

Democratising cultural activity

Generative AI will democratise cultural activity, enabling more people to make more art, but creatives need to ensure they benefit and retain the fruits of their labour. The deal reached between the Writers Guild of America (WGA) and the Alliance of Motion Picture and Television Producers is significant. Rather than fighting the tech by banning it, it sets up a framework that leaves control of AI in the hands of the creatives, not the management who might use it to replace them.

One welcoming aspect of the deal is the implicit recognition by creatives that AI can be useful. Amid the hype, the truth is AI is a 'stochastic parrot’ – very good at working out the next entry on a string of information, but unable (yet) to make great art on its own.
It works by reading what’s on the internet, that’s all it knows. For example, ChatGPT is quite good at multiplying by 9/5 and adding 32, but terrible at near identical task of multiplying by 7/5 and adding 31. Why? Because the former sum is often used to convert from Celsius to Fahrenheit and the latter sum doesn’t appear on the internet at all.

Similarly, when generative AI is used creatively, it can only draw on what it knows – what’s on the internet. The upshot is, as AI writer Vauhini Vara admits, it generates clunky writing, or reduces the world to biased stereotypes or mathematically average stories. As the New Yorker puts it: “Chat GPT is a blurry JPEG of the web”.

Left to its own devices, AI will strip all the subjectivity, nuance and creativity from its output and over time spiral up its own fundament. We’ll end up living in a world where every wall in every room has been painted magnolia. The point is made more academically in this paper from Cornell University: “The value of data collected about genuine human interactions with systems will be increasingly valuable in the presence of content generated by LLMs in data crawled from the Internet.”

Challenges ahead

There are other issues to address. Democratising cultural creation means vast amounts of new art. Perhaps it doesn’t matter that an estimated four million tracks on Spotify have never been played – Forgotify is here to help you find them. Music streaming platforms have evolved to deal with this. But it’s created problems for online sci-fi fan fiction sites. One, Clarkesworld, has had to shut down as a result of being deluged with terrible stories generated by people using ChatGPT.

There’s a fight looming too about copyright. Not only are creatives rightly concerned that tech companies have trained their algorithms on material scraped from them without permission, and are making money from it, but who owns the copyright for material created by generative AI?

As regulators begin to – belatedly – address the issues, there are many things they need to think about, not least the tendency of new network technologies to concentrate power in the hands of an increasingly small number of actors.

So, it’s great to see people in the creative and cultural sector beginning to think seriously about how to engage carefully with generative AI. The BBC has recently announced its three principles for working with AI. People in the Cultural Heritage sector are also thinking hard about how to use AI. A collaborative group AI in Cultural Heritage has quickly amassed nearly 200 members.

Elsewhere, software platform providers Axiell have been dipping their toe into AI, working with the Britten Pears archive to “uncover hidden truths behind one of Britain’s greatest composers”. And at Historic England, Interim Content, Marketing & Digital Director, Richard Worthington is using AI to generate audience-facing material to better present heritage content to local audiences, using data and content from different silos of data. The aim is to generate pages that boost audience engagement and encourage exploration.

But Worthington says that, while AI can produce hundreds of pages in the time it would take a human to create a handful, humans are still necessary. AI, he says, “added more context, but was often wrong, making assumptions and including stereotype. [It’s like] having a junior member of staff… you need to check their output and hope they learn over time. AI doesn’t replace people, you still need that pipeline of talent.”

It’s not too late

Generative AI is here to stay – and the arts, culture and heritage sectors have a powerful role to play in its development. We’re the people making new stuff. As Amy Adams, Collections Information & Access Manager at the National Museum of the Royal Navy and one of the founders of the AI in Cultural Heritage points out, the sector’s values will be essential in shaping the culture around AI and informing the coming regulation. “In some ways the horse has already bolted… AI technologies are already embedded in daily life … your Netflix algorithm and your Spotify and what your kids are asking Alexa about their homework projects,” she says.

“It's not too late, but… as historians and archivists and curators, it’s our bread and butter to think about these questions and how we present history. It's ingrained in our professions, so I think we can offer something to the technology. We can help take these developments and ground them in an ethical way that a profit organisation doesn't necessarily have to.”

Richard Leeming is a consultant with The Audience Agency.
theaudienceagency.org
@audienceagents | @richard_leeming