What if a future president, faced with a sudden global crisis, natural disaster, or economic meltdown, could instantly draw on the wisdom and lessons learned by previous administrations? And what if those experiences were easily accessed across the full range of policy issues and institutional challenges, no matter how vast or arcane?

That’s precisely the kind of research that new artificial intelligence (AI) tools can begin to tackle, as the Miller Center aims the power of AI on its vast presidential studies research collections.

As the Center nears its 50th anniversary, AI technologies offer innovative opportunities. AI has the potential to make the Center’s presidential studies data more accessible to policymakers and the public and to project the Center’s mission and scholarship into the future.

AI has the potential to make the Center’s presidential studies data more accessible to policymakers and the public and to project the Center’s mission and scholarship into the future

Miles Efron, the Center’s assistant director of information technology, is exploring a variety of AI projects that the Center could undertake in the near and long terms by evaluating their difficulties, risks, costs, and benefits. Efron has been educating the Center’s leadership and scholars on the critical background and vocabulary for understanding AI, as well as what it will take to unlock key Miller Center data holdings for AI training and deployment. Efron has worked closely with Marc Selverstone, the Center’s director of presidential studies and co-chair of the Presidential Recordings Program.

Efron has a PhD in information science from the University of North Carolina and is uniquely positioned to spearhead the Center’s AI efforts. Before joining the Center, he researched large language models and search engines and taught computer and information science at the University of Texas and the University of Illinois.

In exploring near-term AI projects, Efron has been examining how to expand the search engine index on millercenter.org to improve the quality of search results. Equipped with robust AI tools, such as OpenAI’s Whisper audio-to-text model, Center staff may also be able to efficiently transcribe Secret White House tapes and accurately index page content for search purposes.

“It would take a long time to get a human to go through” these transcripts, Efron explained. The Center’s collection of audio and video recordings includes 33,098 White House conversations, 1,050 presidential speeches, and more than 650 YouTube videos of Miller Center’s extensive event catalogue. “Now AI can go through all that text for you and find the data you need,” he said.

"Equipped with robust AI tools, Miller Center staff may be able to efficiently transcribe Secret White House tapes and accurately index page content"

Similarly, scholars may be able to use a large language model (LLM), such as GPT-4, to create a “big machine-readable” index of the more than 500 published Presidential Oral History Program interviews and summarize them, Efron said. This LLM could also be instrumental for indexing the Presidential Recordings Program’s 4,500 annotated transcripts of White House tapes.

A comprehensive index would enable scholars to effectively “search for and identify key themes and developments” in transcripts, said Selverstone, “so that as scholars we can make greater use of them and derive insights we might not have otherwise—because the materials are just so vast.” Furthermore, an index “can help inform not only our understanding of the past but also trend lines in matters of public policy that affect us today.”

To address challenges posed by old, noisy White House recordings, the Center also is exploring building an audio classification tool that assists scholars with transcribing inaudible passages and identifying speakers.

While undertaking simpler AI endeavors, the Center could tackle more costly, labor-intensive AI projects in the long term. These might include building a conversational system to support wide-ranging, interactive queries into presidential history and politics, drawing on the Miller Center’s unique data to formulate answers.

Efron’s exploration of AI technologies considers the different audiences interested in the Miller Center’s scholarship. AI can provide a framework to help scholars go deep into data, he said, while helping people who aren’t as familiar—whether White House officials, journalists, or students—get a better sense of the Center’s resources.