Latter-day Saint Thought

AI and Gospel Music, and a Public Service Announcement

by  •  • 9 Comments

Note: None of this is an April Fool’s Joke, it just happens to be the day we had a spot available in the queue.

So far the three main AI use cases that have achieved liftoff are Large Language Models, text-to-image, and translation (Supposedly OpenAI has achieved text-to-video that is so good that multimillion dollar movie production investments are being cancelled. Still, for some reason Open AI has not actually released “Sora” to the public, so until we can play around with it it’s hard to know what to make of the hype). 

However, text-to-music has just had its breakout moment. Previous AI-generated music was short and consisted of a series of extremely formulaic pastiches, but this latest model by Suno has achieved breakout, and AI junkies have spent the better part of this week making Viking saga songs about their cats. 

Being a non-music junkie, I feel like 90% of the music content put out by stars basically sounds the same, with 10% of them being the mind worm hits that we all know. My take is that Suno is pretty good at generating the 90% in the style you want. In principle it’s not supposed to let you replicate styles based on particular musicians, but evidently it’s pretty easy to get past the safeguards. 

So what does this mean vis-a-vis the Church? The people I’ve seen trying it out in Latter-day Saint land haven’t had the greatest luck getting it to replicate, say, Mormon Tabernacle Choir style, and it’s clear that Latter-day Saint music was not a big part of its training corpus. Of course, not all gospel-related music has to adopt that style, but as of now it can’t produce the kind of general conference-type tone that our ears are habituated to. Of course, we all know that will change within minutes. 

However, we are a very top-down organization, with the “canonical” (for the next few decades at least) word on music being soon forthcoming with the new hymnal, so I doubt that even quite good AI-generated music will have much of an impact on the day-to-day in the Church. However, looking ahead a little the forthcoming ability to combine and synthesize styles will make it quite a bit easier to generate new ideas for different renditions, and background and other music for gospel-related content producers will exponentially diversify as, like with art, we are no longer beholden to a handful of prominent styles and works that people keep recycling. 

PS public service announcement: Open AI will be releasing voice widespread voice cloning software in the next little bit–it’s been available for a while but it will soon be available to basically everybody, so anybody will be able to send you a message in your mom’s voice asking for you to wire money to an Algerian bank account. So now is the time to have conversations with your family members, especially older ones, about phone calls in your voice not necessarily being you.

9 comments for “AI and Gospel Music, and a Public Service Announcement”

  1. I am not too excited about any of these AI models honestly. They do create something that can be used to hopefully replace PowerPoint and mindless writing and dumb art.

    As a musician no doubt you can get the approximation of music composition by AI model, but it will be forgettable as most mindless human composition without the fun of doing it yourself. Do doubt it can make some primary and drone like hymns for church too. But none of that is why we do music or art.

  2. I agree with Brian G. I’m a musician as well. I saw a great meme on facebook yesterday that expressed my feelings. It said, in essence, that we’re going the wrong direction with AI. It said, “I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.”

    It’s the human contribution that matters. Many thousands of new hymns were submitted for the hymnal, and those represent the creative spirit of dedicated, faithful church members with an inner longing to express their testimonies through poetry and music. AI can never capture that.

  3. That makes sense for the creators, but what about those of us are consumers without being creators? If you can show that people are no better at guessing what is AI art versus non-AI art, then AI art provides a net benefitto the consumers from having so many more options at scale. Plus it can democratize the production process for people who, say, don’t have the manual dexterity to draw but can visualize something and use their language skills to describe the scene.

    Also, in regards to AI doing laundry and dishes, the general consensus is that we’ll have just that in about five more years. I was also surprised that we got art before our Jetson’s Rosie to help us around the house, but that’s where are.

  4. Think of it this way. We could just play a recording of the tabernacle choir or professional singers playing hymns. Or have player pianos at church. Instead we get people to do it – why? Because it is fun to do and the interaction between learning and trying and doing it together as a community is worship.

    I love listening to recorded music. I don’t play near as well and there are songs I can’t play. I like to draw but I will never be able to draw or paint like a professional artist. But I still play and draw. I could just watch other people dance – like I would a ballet or ballroom dance competition.

    You can buy cheap and mass produced art and there is a place for that that AI may fill, but what a boring capitalist way of thinking about production and consumption of art. You could make an ai video of a baseball game that would closely approximate the real thing. But is that interesting? I would still rather watch the cardinals play than an AI approximation of the game.

    Is there a place for using AI for writing and art, probably but someone will need to think of creative ways to use it before I think it replaces art or music. Art or music is interesting because we as humans do it. It is a tool that needs to be used for something other than imitation before I am
    Interested in it.

    At work I am on some teams using machine learning and AI for data analysis and there are some interesting applications but many times even there a simpler solution like a linear regression is the best model.

  5. So, it turns out that “an AI video of a baseball game” or a rough equivalent already exists in the form of sports video games, and they’re consistently some of the industries top sellers. The People Have Spoken, and what they say is that they like throwing a football around and watching pro football and playing Madden NFL. Not necessarily all of them, but there’s a lot of overlap. Lots of people who can’t play in the NFL are looking for ways for personal involvement deeper than watching TV, and video games offer some of it. I mean, I assume they offer it, since I’m not interested in the NFL, but if I were that’s what I’d be looking for.

    AI is going to offer some different things at different levels. Pros are already using the Photoshop integrations. For the rest of us, there are new ways to access art that are more interactive than simply consuming what others create. If you don’t like the Arnold Friberg illustrations of the Book of Mormon, make your own! They’ll probably be terrible, but no one else has to see them, and maybe they will help you look at the Book of Mormon from a new angle. If you’ve always wanted to hear “In Our Lovely Deseret” as Gregorian chant, soon you’ll likely be able to do just that. I wouldn’t want to sing it in sacrament meeting, but there may be something in between private use and congregational use where AI may turn out to be useful. Some ideas will be useful (like using Google Translate to access things you’d have no chance of reading otherwise) and some will be thoroughly terrible (like talking into a phone app for automatic translation in a foreign language instead of learning a language in the MTC). It will take a while to sort out what’s useful and what’s not.

  7. Okay, those are hilarious. They’re not remotely Gregorian chant–Gregorian chant has no harmony. They could almost pass as madrigals, but some of the harmony is too modern. Sounds like the model got trained on lots of Chanticleer and The King’s Singers.

    I’m struggling to identify the use case here. If you just want music to listen to, there’s SO much recorded music out there, and such a huge variety. You really have to be looking for something very specific for there to be a gap AI can fill.

  8. I just created four versions of Come Listen to Prophets Voice in Bluegrass style. I really like Bluegrass and I’m trying to decide which version I like the best. I’m leaning to 1 and 4.

    1. https://app.suno.ai/song/1697af6f-f28e-45f8-ade3-69cf9a8c526a
    2. https://app.suno.ai/song/5bbd496f-6f79-44fc-ab96-b1192ab51b53
    3. https://app.suno.ai/song/e255dd25-0f0a-4b31-81f9-1af3c8c5e820
    4. https://app.suno.ai/song/60a3e82f-3d6a-43b4-848e-696d899242c6

    The variety is impressive, I think. And it was very quick. I love music and play guitar badly – I could spend/waste hours on this site!

Comments are closed.