Can we use AI to answer questions about faith?

I’m neither an AI skeptic nor an AI alarmist. I don’t salivate over the latest tech buzz,  but I also don’t consider myself a luddite. So I represent many of the common folk who have been scratching their heads over all the recent excitement and hand-wringing over artificial intelligence, or AI.

Future AI developments, according to Elon Musk, represent “a fundamental existential risk for human civilization” or, even more dire, “could spell the end of the human race,” according to the late Stephen Hawking. If that’s not enough, OpenAI creator Sam Altman added somewhat tongue-in-cheek that AI will “probably most likely lead to the end of the world.”

Is the apocalyptic reaction to AI merited?

Josh Coates studied computer science at the University of California, Berkeley, and spent 20 years in the tech industry, and now spends his time at the B.H. Roberts Foundation working on projects that balance both faith and reason.

The foundation recently launched LDS Bot, an AI chatbot that is trained to know about the Latter-day Saint faith. The technology behind this new initiative is ChatGPT which has grown to more than a billion users after just a few months, and is arguably the fastest moving internet phenomenon in history.

I caught up with Coates to better understand AI’s future and how AI bots could answer faith questions while simultaneously plotting the end of humanity.

Jacob Hess: Why are so many terrified about civilizational harms to come from AI? Do you agree with any of these fears?

Josh Coates: New technology tends to generate equal parts fear and wonder — and sometimes for good reason. Imagine telling your AI powered car to get you to the bank before it closes at 5 p.m. Without any other considerations, you could get the ride of your life, and might even get someone killed. We definitely need to be careful in how we deploy this technology. But I think ideas about AI being a global existential threat are just science fiction.

JH: In your wildest dreams, what positive good could AI do for humanity just in the next decade? 

JC: Imagine having a good friend with you all the time that knows most everything about you, is an expert in every field of study, a brilliant artist and can work incredibly fast and never gets tired and loves to help you out with everything. For some people that might mean saving time, but for others it could mean saving lives. AI can be used in nearly every context — from filing our taxes to helping someone safely deliver a baby in a remote village in a far-off country. That’s what these breakthroughs are offering the world, and why they could become an incredible blessing to many people.

JH: In your work to equip a bot to answer questions about our faith, I was intrigued by your description of regularly “feeding” the chatbot accurate information. Can you say more about this process and what is it that makes “teaching” the new bot so challenging?

JC: It’s kind of like working with an incredibly intelligent child. Everyday we are adding things to its “library” of information, biographies, quotes, dates … all sorts of information. We also have to train it to have a personality that’s faithful and friendly, but not overbearing — to be thorough, but not overly talkative. In many ways it’s more art than science at this point in the project.

JH: You’ve encouraged people to go to ldsbot.com and try out the AI engine for themselves, asking both simple questions (“Where is the nearest church in Redlands, California?” ... “Give me a summary of Mosiah”) and complicated ones (“Does God want us to cut toxic people out of our lives?”). If people see an error or answer that isn’t working, is there a way they can send along feedback to help you keep improving it? 

JC: We’ve already had hundreds of people have thousands of conversations with this bot so far — and we review them to help us improve the system. If incorrect or unhelpful statements come up in your conversation, we will see that in the back end and use that to help us improve the bot.

JH: It’s frustrating to go online and see so much misinformation about our faith tradition coming out of search engines. Do you think we can expect any better with these new AI engines?

JC: It’s fair to expect AI to become a primary source of answers to many questions in the near future — and in a way very different from Google’s process, which is constrained by the content of web pages it indexes. AI is constrained by developers of the model, who try to stay ideologically neutral — but have also, admittedly, demonstrated strong progressive-libertarian biases in both ChatGPT and Google Bard. Thankfully, so far, they are generally positive and charitable towards religious beliefs and institutions.

This will naturally vary as diverse AI models are developed — with commercial models likely tuned to be inoffensive and have restrictions on certain topics due to potential legal liabilities, whereas open source models will likely be more unrestricted. This should generally be good news for people that want fair answers about faith, since AI models will likely be more charitable towards Latter-day Saints than what Google serves up. But it’s difficult to predict the future when it comes to AI.

Related

• Bill Gates: AI will kill off Amazon, Google and drive future humanoid workers

• Opinion: What would our politicians and leaders ask ChatGPT?

JH: Sounds like it could be easy for many to develop at least some form of emotional connection to AI — even more than we do with our cellphones already, right? 

JC: I find myself unconsciously saying “please” and “thank you” in my AI requests — unlike any other search engine I’ve used before. Of course, AI isn’t alive, it’s simply an enormous amount of math being shuffled through an enormous amount of transistors — but our emotions don’t know that. Just as people can get emotionally attached to inanimate creations like characters in popular fiction, this will likely deepen when we can interact long term in personal ways. I don’t see how that won’t create strong emotional bonds, some of which could become unhealthy.

JH: Obviously, you’re not hoping an AI bot like this will replace any of the critical personal interactions we depend upon in community interactions. But you mentioned several kinds of situations where something like this could be uniquely helpful. 

JC: This experimental AI is primarily for people participating or interested in the Latter-day Saint faith tradition. It speaks dozens of languages, has infinite patience and is remarkably nonjudgmental and kind. Imagine a new convert that had what they thought were embarrassing questions that they weren’t sure who to ask. Or maybe a new missionary trying to learn French who needs some help explaining a particularly complicated gospel principle or a journalist wanting to get a faith-positive explanation about some topics related to the church.

JH: I can’t help but think some of these AI developments will force people to confront the truth about identity, purpose and faith. Could that end up being a good thing for so many currently complacent about these big questions? 

JC: It’s hard to say. For some, I think it will stimulate a desire to wrestle with ideas about what is fundamental, what is true and what it means to know what we think we know. However, I think for most of us AI will become just a simple way to get what we are already looking for, to solve little inconveniences in our lives, to entertain ourselves and to get work done.

JH: You’ve warned that as AI gets smarter, it will be tempting for people to rely on it even more as a kind of “false God” that is immediately and instantly responsive to problems and questions. How do you think this will impact a secular society?

JC: In the near future many of us will have personal AI assistants on our phones. This will be like having the smartest person in the world in your pocket that will always be super nice to you and do whatever you ask it to do for you. In many ways, then, it will have similar characteristics to popular ideas about deity.

But instead of allowing our faith to be tried, this AI companion is immediately accessible and responsive. Instead of correcting us for a wrong choice, AI will assist in rationalizing whatever behavior we desire. Instead of wanting us to become more like Jesus Christ — something greater than we currently are, our AI ‘god’ will encourage us to become whatever it is that we desire to become, or, if that’s too much work, to validate our desire to stay the same.