AI in culture: where do we stand as a sector

Experiment: AI as a tool in oral history

Erik van Vliet from Erfgoed Gelderland explored how AI can add value to oral history: capturing personal memories through interviews. This preserves stories, emotions, and voices that are not recorded in other sources.

In each step of oral history research, AI can simplify or speed up the work:

1. In the historical preliminary research, for example, NotebookLM is a useful tool for analyzing or summarizing existing sources.
2. ChatGPT can assist in preparing the interview by generating good questions based on your preliminary research.
3. The recorded interview can then be transcribed: converting spoken text into written text. Several transcription tools are available that can do this much faster than a human, including Whisper and the transcription features of Microsoft Teams. Erik shared his prototype on the DEN community platform and via Github.
4. When converting the transcription into a public product, such as an article or infographic, ChatGPT or a similar AI tool can be very useful. There are also numerous tools available for video editing.
5. Although this still needs further experimentation, AI can also support data archiving according to Linked Open Standards.

Can you entrust private matters to an AI tool?

Of course, there are also caveats to using AI for oral history. Can you entrust private matters to an AI tool? Additionally, tools can make mistakes, such as providing incorrect information or transcribing inaccurately. There is also a proliferation of tools, which increasingly require payment. Moreover, technology often suffers from emotion and context blindness: AI can sense what is being said but not how. Finally, stereotyping is a risk with AI tools, whether in interpreting interviews or formulating questions.

The quality of the tools is improving, but it is not yet good enough to take over all tasks. Humans remain indispensable throughout the process. It is important to use your common sense and critically evaluate what AI can and cannot do.

AI in your own organization: user groups, training programs, and guidelines

Jiska Jacobs is the initiator of AI implementation at TivoliVredenburg. Their approach includes the following:

1. The Utrecht venue has an AI user group with fourteen colleagues from various departments who explore new tools together, discuss dilemmas, share use cases, and experiment collaboratively.
2. Through lunch lectures, strategy sessions, and workshops, the entire organization is involved and trained in AI. Topics include ethical and responsible use of AI, effective prompting, and the possibilities of Microsoft Copilot. The training program is conducted by Bureau Reuring. TivoliVredenburg received a subsidy for this via WerktuigPPO.
3. TivoliVredenburg has established AI guidelines, such as sharing sensitive data responsibly, being transparent about your own AI use, and respecting intellectual property. The guidelines are shared on the DEN Community platform.

AI in theater: reflecting on the role of technology in creative processes

Makers are also experimenting with AI's possibilities. Marijn de Jong from performance collective URLAND knows this better than anyone. The collective used AI to create the performance Formerly Known As, in which exhibition makers ask ChatGPT to write a performance for them. The subject of the performance: the creative process of a production using AI. The result is a collaboration between humans and machines, with scripts generated by ChatGPT and cloned voices. The two main insights from this process:

1. The creative process posed quite a few challenges. For example, the makers sometimes didn’t know who had come up with what (the AI or the makers), partly because ChatGPT used the plays URLAND had previously created while writing the script. They also had to consider the rights of others' works they wanted to include in the performance.
2. Both the audience and the makers reflected on the performance. Is staging a production still worthwhile if technology takes such a significant part in the creative process? Does the audience actually believe the work was entirely written by AI? And how does that affect the experience? The performance prompted visitors and creators to think about the changing meaning of art in the digital age.

AI in museums: engaging youth with interactive storytelling

Nationaal Monument Kamp Amersfoort aims to better highlight experiential stories about World War II, especially for young people. Thus, director Micha Bruinsvels decided to develop a digital chatbot. Visitors can use it to converse with guards and prisoners of the camp. Four insights from this application:

1. The chatbot's storyline is inspired by the experiences of former prisoners. This authentic component was essential for Kamp Amersfoort. With sensitive topics like this, you absolutely don’t want to miss the mark.
2. Determining the right approach for the chatbot was not an overnight task. Kamp Amersfoort created multiple prototypes, which they repeatedly tested with the target audience. This helped developers and staff better understand what works well. For example, the story no longer takes place in the present but in the past, and now features multiple characters, creating a more dynamic experience. Changing backgrounds were also added to keep young viewers engaged.
3. Through this form of interactive storytelling, Kamp Amersfoort connects the situation of the past with the present. The conversations address sowing division: during World War II, different prisoners were pitted against each other. By making young people aware of this, they can not only empathize with the past but also prepare for a future where similar dynamics might recur.
4. Kamp Amersfoort built on earlier experiments by Museum de Heksenwaag. By working with the same developer (ImproVive) and applying for subsidies from Cultuurloket DigitALL, they made this project a reality.

Summary

• AI is a rapidly growing theme. Where we barely two years ago asked how we want to relate to AI, today we see many useful applications: as a tool in oral history research, as a creative sparring partner for makers, and as audience interaction in museums.
• Funds like Cultuurloket DigitALL can contribute to realizing your AI project. With the WerktuigPPO scheme, you can partially reimburse training costs.
• Most cultural organizations choose a responsible use of AI with an AI policy. Use the AI policy roadmap to make choices that fit your organization.