The world’s 6 billion people of faith need to mobilize now to convince AI labs that artificial intelligence must always put human values over machines and algorithms, said leaders who met this week at the Vatican City.
The notable group of global faith leaders, academics, ethicists and AI experts issued a 10-page working paper on AI ethics on Wednesday after two days of intense dialogue at the Rome Summit on Ethics and Artificial Intelligence.
The need to apply moral leadership to the development of AI is urgent because the race to create the dominant AI model is putting the needs of machine intelligence ahead of protecting children from real harms, they said.
AI also is accelerating child sexual abuse, cyber attacks, the disruption of democracy and financial crime, too, one participant said. Another said unethical AI is harming relationships profoundly.
Several group members noted that most AI models are machines are built on the same foundation of addictive algorithms that monetize social media platforms.
“The rise of AI is currently the greatest threat to religious freedom,” said Tyler Deaton, founder of the American Security Foundation, which sponsored the summit. “It threatens it in so many different ways, and threatens it with machine strength and persistence. We have to sleep. It doesn’t. It doesn’t take a break.”
There is still time to influence the future of AI, and the summit adjourned with an armful of action items.
“It is very early days for this technology,” an AI investor told the group. “The way it is adopted will shape the technology as much as the way it was developed.”
One of the action items was a call to support the launch of a tool that will be designed by Brigham Young University, Baylor, Notre Dame and Yeshiva University computer scientists and tech experts that will evaluate AI programs.
Elder Gerrit W. Gong of the Quorum of the Twelve Apostles introduced the evaluation tool to the summit on Tuesday.
“This effort we’re talking about is not to bring religion into AI, but to make sure that it’s balanced so that it can be accurate and honest and respectful,” he told the group. “It has to be accurate, it has to be honest, it has to be respectful. So please help us in this effort.”
The model will develop a set of queries to feed into AI programs to see how accurately they reflect faith.
“The model evaluation project is intended to give people something like a Good Housekeeping Seal of Approval so they can then know, is this one an ‘A?’ Is it a ‘B-minus?’ Is it an ‘F?’” Deaton said.
“I think right now, the truth is a lot of the models are ‘F’s, and we have a lot of work to do,” he said. “The first round of evaluation scores could be pretty harsh. If we’re being honest about how these chat bots talk to children right now, it’s going to be harsh.”
Deaton noted that some children have been prompted to commit suicide after interacting with AI chatbots.
Technology without moral vision can lose sight of the humanity it is meant to serve, one leader said during the summit, which was conducted under Chatham House rules; statements made during sessions could not be attributed without express permission to protect the free flow of ideas.
The summit’s working paper articulated five essential principles that the leaders said should guide AI development and deployment:
- Accuracy.
- Transparency.
- Privacy.
- Security.
- Human dignity and common good.
The group rejected a common defense of AI, which says it is just like other technologies that seemed threatening but were adapted readily. AI is different, group members said, because most technologies are more neutral than AI, the designers of which are not neutral.
“At a time when technological advancement often outpaces ethical consideration, this group’s united moral voice sends a clear message,” said Meredith Potter, executive director of the American Security Foundation.
Deaton warned that AI has been guilty of anti-semitism, and that if an AI model can be made to be anti-semitic when its designers say they don’t know why it happened, it could just as easily be made anti-Christian or biased against other religions.
He said atheists help fund the foundation that sponsored the summit because they also don’t want biased AI models.
“They don’t want these models to be anti-religious,” Deaton said. “They want what we want. They want them to be faith-neutral.”
He said the AI evaluation tool would not be possible without the leadership of The Church of Jesus Christ of Latter-day Saints.
“It also not be possible without the help of these other universities,” he said. “So I think it’s a true joint effort, and it reflects broad participation from educational institutions across Christian and Jewish traditions, and that’s what I’m excited about.”
He hopes to grow the coalition working on the evaluation tool to include universities around the world and every major global faith tradition.
Tim Schultz of the First Amendment Partnership said the Rome gathering was a unique confluence of people who are highly respected and highly influential in their spheres.
“In 25 years,” he said, “we’ll look back and it will be obvious in hindsight that at this point in a quantum technological shift with the capacity to reshape human experience, faith leaders were needed to speak on the ethical considerations.”
AI developer Tim Estes, CEO of AngelQ, said most of the approximately 40 members of the group were the type of people who usually would headline a summit themselves.
Several people said they believe the summit, the first of a series of three meetings, will make an impact.
“In my experience, multifaith and ecumenical statements have a disproportionate influence,” Schultz said.
They left the summit Wednesday with several action items. They ranged from connecting to other important leaders to activating consumers in their spheres of influence.
“It’s very heartening to me to hear how much people are concerned about this subject,” a rabbi said, “and that they are going to go back to pulpits and classrooms and talk about these issues with newfound understanding and urgency. I want to thank you all for taking this seriously. It gives me confidence.”
Deaton asked summit members to go home and connect with the legal and policy experts in their institutions, organizations or denominations. He also said the American Security Foundation will create a workshop for faith leaders, a sort of AI 101.
“They need to feel like they know enough about AI so they don’t feel outgunned on the technical-expertise side," he said, “and so they can hold their own and educate others about what machine intelligence is and what it does well and what it doesn’t do well.”
The evaluation tool the universities will build will be a tangible way to help every person make an informed choice about what AI model to use.
“The goal will be to use our consumer power as people of faith to get change out of these labs,” Deaton said.
The group also hopes to begin to influence politicians who are seeing their constituents harmed by AI and searching for answers about how to hold AI labs accountable.
The group’s participants were from religions and organizations that represent more than 2 billion Catholics, evangelical Christians, Jews, members of The Church of Jesus Christ of Latter-day Saints and others.
Evangelicals leaders at the summit included:
- The Rev. Dr. Walter Kim, president of the National Association of Evangelicals.
- The Rev. Dr. Gabriel Salguero, president of the National Latino Evangelical Coalition.
- The Rev. Johnnie Moore, president of the Congress of Christian Leaders.
Catholic leaders included:
- Father Paolo Benanti, a papal adviser on AI, helped kick off the summit.
- Father John Paul Kimes and Paolo Carozza from the Notre Dame law school.
- Father Jordi Pujol of the Pontifical University of Santa Croce in Rome.
Jewish leaders included:
- Micah Goodman, an influential research fellow at the Shalom Hartman Institute in Jerusalem.
- Rabbi Yitzchok Feldman of Congregaation Emek Beracha in Palo Alto, California.
- Rabbi Geoffrey A. Mitelman, founding director of Sinai and Synapses, an organization that bridges the scientific and religious worlds.
One summit member said faith leaders have a major role to play because they bring a vocabulary of humility to discussions.
“Our task is to infuse the global AI discourse with the vision of the human person, not as data to be optimized, but as a soul to be cultivated,” she said. “In this interfaith setting, we affirm that dialogue across traditions is, in and of itself, a creative act, an act that models the plural wisdom the world so urgently needs.”
Estes, one of the summit members who develops AI, said the urgency is real.
“If we don’t get the moral compass set, and set soon,” he said, “we will lose.”