It seems like everybody has been talking about A.I. for a while now, but in the last bit here Christians (especially online-active Christian authors) have been responding to The Gospel Coalition’s “Line in the Sand” on their use of A.I. Yesterday, Tim Challies provided a helpful summary and reflection on the discussion here.
What A.I. Is
I pretty much agree with everything Tim says, but wanted to flag one reminder that I think is helpful to keep in mind: A.I. is not a “thing.” A.I. is an aggregate, a conglomerate, of other things, and those things all come from people.
This might need some explaining. The current batch of powerful A.I. tools like ChatGPT and Claude are, under the hood, large language models (LLMs). They have been fed a huge amount of text and have analyzed it all to estimate what words tend to follow other words. This is why some people have referred to A.I. as little more than fancy autocomplete.
We’ve had autocomplete on our phones and computers for a long time: you type in “Bob hung up the…” and the computer suggests “phone” because, given on the data it was trained on, it predicts that to be the most likely word to come next. LLMs can be thought of as autocomplete on steroids. Lots of steroids. They’ve been fed a mind-boggling amount of data, their prediction models have been tweaked and souped up by huge teams of really smart people, and they have access to powerful computer infrastructure that didn’t seem possible in the not-too-distant past.
There is a lot more going on: A.I. tools also employ neural networks modelled after the human brain, and there are aspects of how LLMs work that seem to baffle even their creators. Still, it’s helpful to remember that LLMs are essentially prediction tools, and when they generate what appears to be intelligent language, we’re actually reading a series of statistical guesses made by a computer program that is really good at estimating what word should come next.
And it was trained in all of that guessing by reams and reams of words written by real people. That’s right: in behind A.I. is the work of humans, real people whose real words form the basis for the LLMs’ prediction models. This is why computer scientist Jaron Lanier says, provocatively, “There is no A.I.” A.I., in his way of framing things, is “a collaboration of people instead of a new entity.” You can listen to a 9-ish minute interview with Lanier here where he explains this idea in a way that is both helpful and compelling.
I think this helps bring a bit of clarity to a key question Tim asked in his article:
Today, individuals and small organizations are identifying AI as a tool that can elevate their writing by providing them with what formerly only institutions could afford—the services of a high-functioning research assistant and editor. The bigger institutions don’t need AI to serve these functions because they have access to funding, and therefore to professionals, who can carry them out… And that brings us to the heart of the issue: Does it matter if this function is carried out by a someone or a something? Is it acceptable for individuals to use AI to narrow the competence gap that institutions enjoy by virtue of their size, or is this giving an app a function that is essentially human?”1https://www.challies.com/articles/tgc-ai-and-the-democratization-of-competence/
If we understand that A.I. is essentially a computer-organized collaboration of people, the distinction between “someone” and “something” shrinks somewhat. When an author uses A.I. to edit an article they wrote, the computer program is estimating how someone would perform that editing based on it’s exposure to the work of many someones.
There’s still a real difference, but it’s not a difference between a human and a machine. It’s a choice between a human and a software-negotiated collaboration of humans. This understanding doesn’t necessarily answer our ethical and practical questions concerning A.I., but it does sharpen them somewhat.
How I Use A.I.
For me, this means that I don’t use A.I. to generate any of my writing—in my sermons, my academic work, or here on this blog. (That em dash you saw there? I’ve been using them for years.) It would be wrong to pass off A.I. generated work as if it were my own, as wrong as plagiarizing another human—because that’s essentially what I’d be doing.
I further agree with Brett McCracken that “writing is a product of deeply human, necessarily inefficient processing, meandering, wrestling, noodling, cogitating, and often suffering.”2https://www.thegospelcoalition.org/article/ai-tgc-missional-opportunity/
The same could be said, perhaps in an intensified way, regarding sermon preparation. Truly, few experiences are richer than standing before my congregation each Sunday to deliver words that I’ve sweat and prayed over all week. I don’t want to preach other people’s words, cobbled together by a computer. It would fundamentally change the experience in a way that has zero appeal to me, besides the obvious ethical issues. I’ve told my church that they should fire me if I ever start using A.I. to generate my sermons, although, honestly, it doesn’t feel like a temptation right now.3Although I do hear the warning of 1 Corinthians 10:12 as I write these words. For similar reasons, I’ve also never used A.I. to write an email or text message. Communication is precious and personal to me, and I’m not comfortable outsourcing it.
But in a variety of other ways, I do use A.I., sometimes extensively. My basic standard is that I’ll use A.I. for anything that I would reasonably ask another person to do. I’ve been a Claude Pro subscriber since March, and here are some of the ways I’ve found it to be well worth the $25 or so I pay each month:
- making meeting minutes. In church committee meetings, rather than tie up someone’s brain power with minute-taking, I record the conversation, transcribe it with MacWhisper, and ask Claude to generate minutes using the transcription. It does a really good job.
- research assistant. Claude is surprisingly effective with scholarly research. I use it many weeks in sermon preparation to test ideas, ask for sources, and synthesize information. Here is an example of an actual chat from my study on Habakkuk 2:4. I don’t trust every word that Claude spits out, but neither do I blindly trust every word written by a human in a commentary. I found and read every source referenced in that chat, which means that it actually helped me read more instead of doing my reading for me. Claude has tracked down sources, tested hypotheses against actual evidence, and given feedback in meaningful ways.
- personal assistant. Claude Cowork takes care of a lot of busywork for me like organizing files, analyzing my schedule, clearing up unused disk space, and other personal admin tasks.
- giving me a head start on projects. I recently used Claude to create discussion guides from a series of teaching transcripts I did. I still edited the end results somewhat, but using A.I. saved me 3-5 hours of work. I also used Claude to summarize a book chapter for a devotional I gave once, and although I told people I’d used A.I., I had to edit it so much to make it sound like me that I don’t think I’ll do that again.
- editing. I don’t actually use Claude for this very much, if at all. But for the past few months I’ve used Apple’s built-in Writing Tools to proofread every sermon manuscript before it gets posted on the website. This is a pretty basic form of A.I., but it usually catches 30-40 grammar or formatting mistakes in each document. (I don’t use the “rewrite” feature at all.)
There’s probably more that could come to mind, but these are some of the big uses. And that’s just for ministry-related tasks! Personally, I’ve used Claude to build three apps, deploy a project server, research a ton of topics, and do a bunch of other work that would have taken me months on my own.
A.I. is a tool, a powerful tool, and a tool for which I thank God: it has been a big help to me personally. Like all powerful tools (think of knives and fire) it has some inherent dangers, but with proper care, I believe it can be used with great benefit. Hopefully these short reflections encourage you to do the same.
