Case study | Exploring the uses of generative AI

The use of AI is not new to the Smithsonian, as many staff have been using these tools for years to accelerate research, enhance the institution’s collections and automate internal processes.

Experts from around the Smithsonian, including my own office of digital transformation, are engaging deeply with AI projects and discussions across our sector, and speaking and publishing on the promise and pitfalls of this technology to transform educational, cultural heritage and scientific organisations.

But with the rise in generative AI – technology that can generate text, images or other media – the use of these tools to collect, organise and disseminate information is becoming increasingly prevalent. It is transforming how the world finds and shares knowledge.

At the Smithsonian, we are exploring the practical uses of generative AI, experimenting with how it might help us engage with public audiences in innovative ways and how it can transform our internal processes to increase efficiency.

We are also exploring a uniquely valuable role the institution might play as a trusted source in the development of AI tools. Given how AI models are built, they require accurate, reliable information on which to learn, to ensure their future outputs are true.

Drawing on more than 175 years of collections, research and scholarship, the Smithsonian can make well-curated data available for AI training purposes, allowing the institution to play an important role in ensuring these models are informed by trusted and accurate information.

Beyond thinking practically about these tools, there are bigger questions to consider when it comes to AI. In a world where machines will soon be able to automate many cognitive tasks, how will society continue to value the power and primacy of human ideas, creativity and lived experience?

AI and Digitisation

To tell the story of the first American in space, the Smithsonian’s National Air and Space Museum in Washington has been utilising AI to produce a 3D digital version of one of the museum’s most iconic objects: astronaut Alan Shepard’s 1961 space suit.

The metallic sheen and texture of the suit made the object challenging to capture in fine detail through the 3D digitisation process. Reflective surfaces do not render well, as technicians take images using a process called photogrammetry, which creates 3D images from photos.

Applying AI technology developed by Adobe Research facilitated the use of photogrammetry and light and laser scans to accurately capture the reflectivity of the suit to produce a detailed “digital twin”.

The digitised 3D version is available online, offering the public a close-up look at a milestone moment in aerospace history.

Museums, archives and libraries are stewards of cultural values, and being human-centred is one of our sector’s core tenets. The Smithsonian is a dedicated keeper of America’s unique knowledge capital – art, science, history and culture. AI systems should amplify these human capabilities, not replace them.

While there are concerns about bias, ethics and safety in some of the biggest AI products, the Smithsonian stands steadfastly for values such as scholarly research, scientific exploration and creative expression. Machines don’t hold values or culture – only people do.

So, while the age of AI stands to boost the work of the Smithsonian, it also makes what the institution provides more precious than ever: a reservoir of knowledge encompassing our nation’s ingenuity, culture, exploration and achievement, and a reliable source of human truths that the public can rely on to navigate our shared future.

Becky Kobberod is the head of digital transformation at the Smithsonian