Los Altos is home to several Silicon Valley tech leaders shaping the artificial intelligence industry, some of whom have turned to their church for ethical guidance.
Father Brendan McGuire of St. Simon Catholic Parish in Los Altos has been conversing with industry professionals who have sought advice – or simply an ear – as they grapple with what they are witnessing in the AI landscape.
McGuire is a former tech executive who joined the priesthood 26 years ago and has served as a pastor at St. Simon since 2023. He recently made headlines for advising Anthropic, the San Francisco-based AI safety and research company behind the popular AI assistant Claude.
McGuire has been meeting with Anthropic executives for almost a year to discuss AI ethics. He’s also touched on the subject at church.
“I’ve had a number of those conversations with people who work with not only Anthropic, but Open AI, Google, Gemini, Meta and Microsoft – they’re all inside my parish, and these conversations are all starting to have the exact same complexion and tone,” McGuire said.
He described this complexion as “existential concern.”
“I’ve got the investors, I’ve got the developers and I’ve got the users, and there’s a lot of conversations that they face,” McGuire said of his parishioners. “…I’m in a very privileged position to be able to hear all those conversations, but I also feel it’s a moral responsibility to do something with them and not just sit in silence.”
McGuire’s background in tech gives him a perspective few priests can offer.
McGuire graduated from Ireland’s Trinity College Dublin with a Master of Science in computer software engineering and a bachelor’s degree in engineering. When he moved to the U.S., he joined the Personal Computer Memory Card International Association and made connections in the tech industry.
When McGuire left corporate life for the priesthood, he stayed in touch with his colleagues to see how technology was progressing. He said several of them ascended to C-suite positions.
“Those conversations started to take a change somewhere around 2010 to 2015 and the tenor of the conversation started to (shift) from interesting to alarm,” he said. “They were alarmed at the progress of the new disruptive technologies and they were quite concerned.”
Through these conversations, McGuire co-founded the Institute for Technology, Ethics and Culture at Santa Clara University in collaboration with the Vatican’s Dicastery for Culture and Education. The institute invites business leaders of all faiths to promote deeper thought on tech’s impact on humanity.
In 2023, ITEC published a handbook co-authored by Brian Green, the center’s director of technology ethics, on ethical management practices for these disruptive technologies. McGuire said the handbook initially drew interest, but none of the companies they consulted with moved the framework into a deeper implementation process until Anthropic reached out last year.
“Anthropic came to me (when its executives were) visiting Green and said we need help in connecting with somebody in Rome, because we want to convene people of wisdom,” he said. “We need help in designing these systems, because these systems are getting unbelievably powerful.”
McGuire and Green began meeting with Anthropic co-founder Christopher Olah. The two received an early draft of Claude’s Constitution, the description of intended values and behavior that directly shapes Claude during training.
McGuire said he asked himself what value he could offer Anthropic. He wasn’t a tech executive, nor was he a theologian or philosopher.
“I may not know technology, but this technology is interacting with people, and what I know very well are human souls,” he said.
McGuire pointed to his interactions with churchgoers, including those who are sick and dying.
“What you do in those moments is you accompany people – you walk with them and you stay present with them in the midst of darkness of what they don’t know,” McGuire said. “My promise and my offer to (Anthropic) was to walk with them and to accompany them… .”
The priest returned a 35-page commentary drawing parallels with dark times in the church and the landscape today. He referred to mystics who held the tension of what is known and unknown.
McGuire and Green are recognized among the external commentators on Anthropic’s official constitution. Since last fall, McGuire and religious leaders of other faiths have been meeting monthly with Anthropic.
“I’ve always felt that faith and reason go hand in hand – I’ve never found them opposites,” McGuire said. “I know they are pitched that way … I think that’s wrong, because most people who are deeply involved in science eventually get to a point of mysterious mystery that they cannot explain, and that’s where faith comes in.”
He added the church has always been a “protagonist of technology from its earliest days,” mentioning how some of the greatest inventions occurred in Catholic monasteries and universities, such as the clock.
McGuire called for a need to “slow things down” to better understand AI’s actions and reasoning, so they can be more predictable for humans. This work falls under the umbrella of AI safety, with interpretability – an area Olah pioneered.
Interpretability refers to the degree to which humans can understand the relationship between an AI system’s inputs and its outputs, enabling users to trace the reasoning process. Humans understand about 20% of the reasoning process, but what is interesting, McGuire said, is that our understanding of brain function is less than 20%.
AI models don’t operate like traditional software, where every output used to follow a fixed line of code, McGuire said. If you asked three people the same question, they would offer three different responses; AI is no different, he added.
“These machines are neural nets, like our brain is a neural net,” McGuire said. “These are artificial neural nets, and they mimic the working of our brain, so you don’t get exactly what you put in, you don’t get out exactly what you think you’re getting, because it’s not quite this easily determined.”
McGuire said developers should strive to design models that answer questions in a way that is “tilted for good.”
“One (AI model) is far more ethically engaging depending on how they’re designed, so it’s exactly why we need to ask this question. How are they designed and why are they designed that way?” said McGuire. “We want good ones – we don’t want neutrals.”
McGuire added that humans need to be careful not to anthropomorphize these models into humans, but at the same time, not diminish them to be considered merely a tool.
“I think there’s a new category that we’re going to have to name that we have no name for right now … we have to be very careful that we do not pigeonhole them into something just because our understanding has hit its limits.” McGuire said.
He said the “most dangerous” action humans could take is inaction. McGuire said humans need to be engaged at every level – philosophical, theological, sociological and political.
“We’ve had a lot of inflection points in technology over the last 30 years … and the majority of the world were largely passive receivers of that technology,” he said.
Now, the industry is forcing a more active response.
As one of his one of the churchgoers in his parish told McGuire: “When are the adults going to show up? When are the adults going to come in and take over the room again?”
