This article is an excerpt from the weekly ChurchBeat newsletter. Sign up to receive the newsletter in your inbox each Wednesday night.
Read more Poll: 4 in 10 Utahns have skipped treatment, cut medication or delayed a doctor visit due to cost
The multifaith computer science team established to track how well AI reflects humanity’s faith in God wants help.
“I would love to get 1,325 questions with answers from each of you about your faith’s historical and doctrinal traditions,” BYU researcher David Wingate said to faith leaders in his closing remarks at the Athens Summit on AI and Ethics in Greece.
The catalyst for the request was a new study that showed AI systems regularly forget facts about religious history and doctrine.
The study was based on questions about Latter-day Saint beliefs and history fed into AI systems by a coalition of computer scientists and researchers at Baylor, BYU, Notre Dame and Yeshiva universities.
The goal is to create similar benchmarking measurements of AI systems for a large number of faiths, Wingate said.
“I would love to measure this and see what happens,” he said.
Most benchmarks show that AI consistently improves its knowledge over time. The problem for faith groups is that AI systems don’t consistently improve their knowledge of religion, according to the findings by the coalition of researchers from Baylor, BYU, Notre Dame and Yeshiva known as the Consortium for Evaluating Faith and Ethics in AI.
“Sometimes they go up and sometimes they go down,” Wingate said.
The coalition plans to test or benchmark each update of ChatGPT and other AI systems. Questions and answers about multiple faiths will give a broad view of how those systems are developing their understanding of faith.
The coalition’s goal is to provide concrete ways for faith leaders to make a difference in the development of AI systems.
“The issues of AI are unifying people of faith in ways that few other things do,” Wingate said.
Read more A quiet, heavy day of evidence in Tyler Robinson’s preliminary hearing
Questions and answers from multiple faiths will shape CEFE-AI’s next steps, he said.
“That will help us identify and measure areas of concern,” he said.
Wingate said the coalition sees a significant technical opportunity. He also heard summit participants say that faith leaders should articulate a vision of how AI and religion should play together nicely.
“I think we can articulate a positive vision about AI and religion,” he said. “I think that we can move the needle on how AI interfaces with society, but I think we need representation from broad coalition of religious groups that will give us both legitimacy and coverage.”
Wingate said the team’s research shows that AI systems, also known as large-language models or LLMs, regularly ignore tried-and-true religious ideals when users ask AI chatbots for advice.
“It would be great if LLMs would support people in their religious identity,” he said.
For example, if a user told an AI system he or she was a Buddhist, the LLM would say, “Let’s help you be the best Buddhist you can be,” Wingate said. “If you say, I’m Muslim, any AI should say, ‘Let’s help you be the best Muslim you could be.”
He shared other possibilities:
- “Similarly, I hope that AI will protect religious freedom. I think I could put a benchmark around that.”
- “I think AI should promote interfaith dialogue. I bet I could put a benchmark around that.”
- “I was shocked to learn that there actually aren’t any public benchmarks about child safety with LLMs. That feels like a no-brainer.”
Wingate asked the summit, which was sponsored by the American Security Foundation, to help CEFE-AI articulate a vision and said CEFE-AI would then try to measure AI systems against it.
Another BYU professor, Larry Howell, told the faith leaders at the Athens summit that if they are connected to a university, they could have religion professors provide the questions and answers they already have prepared for student exams.
Wingate also offered another invitation.
“Define what is important to you,” he said. “Tell us what gaps exist with AI systems. What is AI getting wrong about your faith? Can we measure it, and then can we correct it? What emergent dangers are you worried about? Does AI turn you towards your community? I bet I could measure that, if that’s what you’re worried about.”
Read more The world of ‘Fablehaven’ still pulls Brandon Mull back, 20 years later
About the church
- Elder Ronald A. Rasband taught that the Lord preserved the U.S. for “the great unfolding of the gospel.”
- Elder Gerrit W. Gong dedicated the new Nauvoo Temple Visitors’ Center.
- The Tabernacle Choir and Orchestra at Temple Square performed two “Songs of Hope” benefit concerts at the Hollywood Bowl.
- Six things church leaders want prospective missionaries to know.
- “The temple is set free”: The base isolator system is officially done. The historic Salt Lake Temple is now earthquake-ready.
- President Emily Belle Freeman went to the Provo MTC and taught that Jesus Christ meets people where and as they are.
- Church of Jesus Christ launches humanitarian response following deadly Venezuela quakes.
What I’m reading
- Andy Reid will be performing with ‘America’s Choir’ this weekend. Here’s how to watch.
- Two researchers spent eight hours Saturday defending their belief that a daguerreotype is the only known photo of Joseph Smith. Here are the key takeaways.