Are museums ready for the intelligence test?

Bletchley Park: at the epicentre of a paradigm-altering technology that could change the way we all live? No, it’s not 1943, it’s the 2020s and the artificial intelligence (AI) revolution is officially here.

The world’s first AI Safety Summit was staged at the Buckinghamshire codebreaking museum last November, attracting world leaders such as the US vice-president Kamala Harris and the SpaceX billionaire Elon Musk (who admired the museum's Enigma cipher machine, telling curators it was similar to one in his own private collection).

Taking place just a year after the launch of the game-changing “large language” chatbot, ChatGPT – which can compose poetry and provide complex answers to users’ questions – the summit showed that AI has moved past the conceptual and firmly on to the global political agenda.

AI tends to provoke passionate feelings, both positive and negative. Evangelists of the technology believe it could finally create a long-promised utopia where bots manage all the drudgery in our lives and work, freeing us up to do more interesting things. Pessimists fear it will destroy the employment market, flood the world with misinformation and eventually destroy humanity itself.

The rest of us fall somewhere in between these two extremes: the technology has many exciting implications but there are some clear risks as well (and not just of the Terminator variety).

The urgent task for museums and the wider cultural sector is to work out how to maximise the potential of these powerful new tools, while mitigating against the legal, ethical and practical pitfalls that lie ahead.

Generational shift

“The pace of change over the past 18 months has been phenomenal. We are at the early stages of a huge generational shift in how we think and work,” said Kevin Donnellan, founder of the AI consultancy Expl.AI.nable, at the recent Newsrewired conference.

Putting AI strategies and policies in place is essential for organisations across all sectors. “The best time to write your AI strategy was yesterday,” he said.

Fears about the long-term risks of AI can obscure some of the more mundane threats, added Donnellan. Chief among these is the impact it could have on jobs – a risk that will affect different industries in different ways.

At present, generative AI tools – that is, tools that use machine learning to create new content such as text, images, sound and video – are nowhere near being able to replicate human expertise; the difference between the two is what the Guardian journalist Chris Moran recently described as “good output versus something that looks like good output”.

Bias also remains an intractable problem – and one that tech companies are not properly addressing, according to Donnellan. For example, a prompt for “museum curator at work” in one of the many free AI image-generators turns up a middle-aged white man – the onus is on the user to refine their search for more diverse results.

And such technology will never be able to replicate the human touch of empathy, imagination and creativity. Rather than replacing people’s jobs, AI should be seen as a tool for taking care of some of the “tasks within” a job, said Donnellan. 

Tentative steps

While there is wariness within the sector as a whole, some institutions are experimenting with AI. Last autumn, several professionals launched the AI in Cultural Heritage Group, an informal discussion network open to anyone working in galleries, libraries, archives or museums.

Birmingham Museums and Galleries recently made 500 of its artworks available on the Kultura art platform, which has used ChatGPT to create a chatbot that enables users to ask questions about the paintings.

“It’s supplying some decent replies,” says Linda Spurdle, head of digital at the organisation. But misinformation, another concern about AI technology, remains an issue. “Hallucinations” – the fake answers that the chatbot spontaneously generates when it doesn’t know something – mean it can’t quite be trusted.

“If we wanted to introduce something like this into the museum, we couldn’t say that everything it says is absolutely true,” says Spurdle. “It doesn’t remove the need to have someone there to understand the collections.” 

But Spurdle believes museums shouldn’t let the quest for perfect accuracy be the enemy of the good. If audiences can’t find information they want from the institution “they’ll just go away and Google it or look it up on Wikipedia,” she says – sites where the museum has even less control over quality.

Back-of-house museum work is where AI could prove its worth, with several pilot projects having recently launched to explore how new technology could be used in collections and archives. 

The Museum Data Service (MDS), which launched last November at the Museums Association Conference, is a platform that will pool millions of digitised object records from museums across the UK. The project is working with academics to explore how the platform might use AI, for example to process and clean data or connect records.

“Computer scientists are interested in the [MDS] dataset as a complex, multi-layered data model,” Ross Parry, director of the Institute for Digital Culture at the University of Leicester, told delegates.

“On one hand, it’s messy and inconsistent, with lots of ways of describing things over the decades, but at the same time it’s a community of practice with a unifying endeavour.

“AI will help us to pick out and connect and extrapolate and make extraordinary inferences from individual object records. It will also help us to look at an entire meta collection across multiple institutions.”

The next five years are likely to see huge AI-driven changes in how the sector works. Museums need to get up to speed – you never know when Elon Musk might drop by.

AI project into Benjamin Britten produces intriguing results

The Britten-Pears Foundation cares for the archive of 20th-century composer Benjamin Britten, as well as running two Suffolk venues connected to him and his life partner, Peter Pears.

Last year, it participated in an experimental archive project with collections management company Axiell and AI company Faculty. The project explored how AI could be used to find links between collections in the Britten archive and other repositories of material linked to him, such as the Royal Opera House.

“As we got towards the end of the process, we thought there are probably more interesting things we can do,” says Christopher Hilton, head of archive and library at the Britten-Pears Foundation.

“What we ended up doing was exploring ways in which we could serve up a network of people by seeing where two names often crop up together – those entities are clearly strongly connected, so let’s link them together.”

This change in approach had intriguing results. “There are people in the archive who maybe we don’t know about, because they’re not singers or named artists,” says Hilton. “What they are is connectors. I suspect it was a highly gendered role, with a lot of women who were excluded from paid employment.”

The AI project uncovered the hidden stories of female organisers who played a pivotal but unseen role in the careers of Britten and other artists. “We could do a lot of surfacing the unseen work of women in a profoundly sexist society through something like this,” says Hilton.