The leader of a Christian ministry that built its own artificial intelligence application to guide users in their spiritual journey warns companies developing AI systems need to implement "ethical guidelines" to ensure their technology doesn't harm people.
asserted
technology → build → people
Ligonier Ministries president and CEO Chris Larson says that the true highest intelligence is God, and shared a new tool his group created to help answer questions people my have from a faith-based perspective.
"It's interesting that [President Donald Trump] even referred to artificial intelligence now as 'superintelligence,'" Larson told Fox News Digital.
asserted
Larson → say → superintelligence
"Of course, Christians — and, really, all of us — have access to the 'superintelligece,' the one who has created the universe.
asserted
who → have → universe
The mission of the Florida-based organization, founded by the late Christian theologian R.C. Sproul, is to proclaim, teach and defend the holiness of God "in all its fullness to as many people as possible."
Part of accomplishing that mission was building an app that the organization calls "Grow."
"
asserted
organization → base → that
It's dedicated to helping Christians to better understand the Bible and theology and to go deeper in their faith, put down roots, if you will," Larson said of the new tool.
asserted
Larson → dedicate → tool
"And so this app has 50 years of teaching in it from more than 800 different teachers, of course, grounded in Dr. R.C. Sproul's lifetime of teaching."
asserted
app → have → teaching
The app draws from a library of tens of thousands of pieces of content help users in their faith journey looking to obtain answers to critical questions.
asserted
users → draw → questions
Inside Grow is an "intelligent answer assistant" which can answer questions from users from a faith-based perspective.
asserted
which → answer → perspective
But building an Grow for Christian users required putting a great deal of thought into the ethics of the model, Larson explained.
asserted
Larson → build → model
"Now, it's not trying to pretend it's a person or anything like that.
asserted
it → try → that
It is just trained on our library, and then helping Christians to trust the answers, and [it] will only go so far
asserted
it → train → answers
," he said.
asserted
he → say → ?
"And again, I think that's an important guard as well.
asserted
that → think → ?
Because for us, making sure that we have trusted theological answers and that the intelligent answer assistant is acting with integrity and not going outside of that.
asserted
assistant → make → that
"
When a user pushes the boundaries of the application, it directs them to real human support — a feature most large language model AI tools refuse to implement.
asserted
tools → push → support
"And so we've given it a sandbox to work within, and any time that somebody tries to push it beyond the borders, we're going to deflect or defer to a real human.
asserted
we → give → human
Imagine that," Larson said.
asserted
Larson → imagine → that
For example, if a user discusses suicidality in Grow, the intelligent answer assistant will direct them to the suicide hotline or other real-world resources.
asserted
assistant → discuss → hotline
Over the past several years, there have been a flurry of suicide deaths that victims' family members say AI chatbots were at least partially responsible.
asserted
chatbots → say → deaths
In one case from last year, a teenager named Adam Raine took his own life after months of discussing his suicidal thoughts on ChatGPT.
asserted
teenager → name → ChatGPT
According to a lawsuit from his parents, the chatbot provided him information about methods of suicide, validated aspects of planning the suicide and offered to help him draft a suicide note.
uncertain
him → accord → note
OpenAI, ChatGPT's parent company, disputes the characterization and says the chatbot directed him to crisis resources multiple times.
asserted
chatbot → dispute → resources
"So, we understood that it's important to provide safety guardrails, but also we did not want this intelligent answer assistant to pretend that it was a person," said Larson.
asserted
Larson → understand → guardrails
"We don't want it to take the place of a pastor or a local church elder or even a parent."
asserted
it → want → pastor
Larson said companies developing AI need to consider guardrails for their technology.
"
asserted
companies → say → technology
Anybody developing AI tools today needs to think through ethical guidelines, and that's more of the self-regulation and guardrails that we've been talking about," he said.
asserted
he → develop → that
"I think there's a real concern today because you have great power that's been entrusted to these different companies and they're not really disclosing their worldview," he continued.
asserted
he → think → worldview
"How do they know right from wrong?"
asserted
they → know → wrong
For Ligonier, the absolute truth upon which its AI model is built is the Christian faith.
asserted
model → build → which
"How are you determining what is right and wrong?
asserted
what → determine → ?
As Christians, we believe we have a true answer — a source that we can turn to.
asserted
we → believe → that
"
If you or someone you know is having thoughts of suicide, please contact the National Suicide Prevention Lifeline at 988 or 1-800-273-TALK (8255).
asserted
you → know → 988
The Aporia
