Whether used for communication, navigation, research, entertainment, or social media, many of us cannot imagine living in the twenty-first century without the internet. Yet the explosive growth in artificial intelligence (AI) technologies has the potential to have as much or more of an impact on the world as the internet. New technologies are not without fear, however. With a new awareness of AI being established by news outlets worldwide, many wonder what the implications will be, as they did with past innovations. Will AI replace my job? Will it render irrelevant or nonessential the skills and abilities I have honed over the years? Will it engage young minds with evil and distract them from good? Various responses are possible, from excitement about the future to fear and uncertainty.
Amid all this uncertainty, how is the church to respond? As a pastor who loves learning and expanding my expertise in ministry and technology to discover how they can work together, I see technology as enabling rather than detracting. God gave us minds and bodies to make the best use of. Without the printing press, we might still be stuck without the Bible. Every innovation has advantages—and disadvantages when misused. This is not a fundamental problem with innovation but with human nature apart from Christ. In other words, Cain did not require the invention of modern tools and technologies to take his brother Abel’s life or to live a self-destructive life. All it took was a self-serving, fallen mindset. This awareness can be helpful as we navigate new trends like AI.
You already use AI
Despite the recent rise in awareness and popularity, AI technologies have existed for years. At the risk of drastically oversimplifying it, “machine learning is a branch of artificial intelligence (AI) . . .
which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving accuracy.”1 Simply put, by observing patterns in data, artificial intelligence can begin to predict and extrapolate what should come next.
Have you ever noticed how autocorrect seems to learn your typing style on a smartphone or how ads pop up for something you were just considering? In these and many other ways, AI technologies are used in the twenty-first century, including face unlock technologies, customized search engine results, smart home assistants, navigation apps that calculate traffic and suggest ideal departure times, spam blockers, bank fraud detection services, media streaming services, and so much more.
The rise of smartphones has enabled companies to collect vast amounts of data from our everyday lives. So, if you have ever wondered why an app, social network, or service is free, you are probably the product that allows them to exist. Habit tracking and data collection make companies money when sold to others who can use it. Marketers create ads designed to reach people with target habits and interests.
The big difference in recent years is that AI technologies are now being trained on massive volumes of data to see how well they can perform new functions. For example, image generation technologies like Midjourney, Dall-E, or Stable Diffusion are based on machine learning that has analyzed so much visual data that it can predictively generate completely original, new images by starting with random noise and then slowly refining it to imitate visuals that are associated with the keywords a user provides. As bad results are voted down and good results are voted up, the technologies are trained to learn and improve.
Text generation technologies like ChatGPT, Google Bard, or Facebook’s LLaMA are also trained on large amounts of data to predictively produce outputs based on prompts entered by a user. The key thing to understand is that because outputs are fundamentally predictive based on data they have been trained on, they may not always produce objectively factual information. This will improve as text-based machine learning models are trained to produce sources to corroborate claims. Nevertheless, just as humans are fallible and need to be fact-checked to ensure reliability, the same applies to AI.
Because machine learning is based on predictive algorithms, text-based AI technologies are good at writing some kinds of emails, ad copy, and other generic administrative needs. When it comes to ministry, however, some cautions are necessary because Christianity has a spiritual component. An AI chatbot might be able to produce a sermon outline that looks like a sermon. Still, it cannot replace the Holy Spirit’s guidance, theological understanding, emotions, and insights into the Christian experience. This quickly became evident after an AI processed the book Steps to Christ and was asked to generate lyrics. Consider the output:
Jesus came to reveal the Father’s love,
A sacrifice made from heaven above.
Through His life, death, and resurrection,
He showed us God’s infinite affection.
He healed the brokenhearted and set captives free,
All to show us God’s love and mercy.
God is love, it’s written everywhere we see,
In every flower, in every tree.
Let His love and truth reign in our hearts,
And from Satan’s grasp, we’ll never depart.
Through faith in Him, we’ll reflect God’s pure light,
And share in the joy of holiness and might.
This initially reads much like something one would expect to read in Steps to Christ, appearing to even quote phrases from it. Still, the italicized line demonstrates an ultimate lack of understanding and theological contextual awareness that can quickly result in outright heresy. With this risk in mind, predictive technologies may be best kept separate from generating content on spiritual or theological themes as long as they are liable to misrepresent even the basics of Christianity and divine revelation.
This caution also extends to anything where authoritativeness is expected. This predictive nature of AI models to mimic things makes them better at imitating content typically regulated by more concrete rules and structures, like computer programming, spreadsheet formulas, common types of letters, and basic structured tasks.
With thousands of developers and companies producing AI solutions, AI is progressing so rapidly that half of these companies could be replaced tomorrow.2 Unfortunately, proposed solutions come with subscriptions. Hopefully, this will change as costs drop.3 Here are some ways I am already using or excited to use AI in ministry.
- Meeting transcription, translation, summarization, and task assignment.4 Most people do not love meetings, much less having to take notes and keep everyone accountable. Several AI solutions have been created to streamline this. Tools that show how long each person has spoken can also be a helpful reminder that brevity is a virtue.
- Basic content creation and refinement. This can include generating blurbs for upcoming event ads or creating sermon or podcast descriptions for online posts;5 AI-generated intro/outro music tracks;6 creative presentation title ideas; logo options for ministries and T-shirts; presentation slides formatted to the content you provide; and grammar checking for correspondence and sermon manuscripts.
- Generating images for flyers, church bulletins, sermon slides, certificates, bookmarks, and more. I recently created custom images for a baby dedication certificate that turned out well. AI can also help produce flyer designs in a fraction of the time.7
- Administrative tasks for communication, transportation, and visitation. In ministry, one of the common challenges I have had is coordinating visitation groups. This involved a spreadsheet of contact details and planning out multiple weekly visitation routes. I would set up a complicated spreadsheet with addresses I had to edit until I could load it all into Google Maps and then use another third-party tool to select the shortest route for three to five addresses. I am excited to use AI integrations with Google Sheets to clean up address data and create formulas to achieve desired outcomes (like listing how many minutes away each address is) and the shortest route for visiting multiple addresses.8
- Streamline planning. Ever struggle trying to decide what recipes to plan for a potluck? AI can suggest meal ideas catered to dietary guidelines and available ingredients. For actual recipes, be sure to use AI that can cite real recipes.9 A recipe that is only 90 percent accurate could end in a distasteful disaster.
- Brainstorming ideas when doing your own writing. I have asked several AI models for synonyms to words or even similar Bible verses or stories on a given topic. They have sometimes pointed me to examples I had not yet considered so that I could look them up and study deeper.10 Outcomes can vary a lot here.
- Improving your understanding of ideas. I engaged AI on the Sabbath’s relevance to Christians today. It helped me better understand common arguments. After forming some personal arguments, I asked the AI to argue for my position. To my surprise, it did a job that could be comparable to an average church member, though admittedly not perfect. I can see this being used in youth ministry as a fun way of challenging young people to search their Bibles to defend their faith better.
AI solutions to look for
- Software to take scanned or handwritten documents and process them to be searched more accurately than through existing optical character recognition (OCR) technologies. This can benefit churches with boxes of records that are hard to search through during audits, not to mention the potential for improving historical publications.
- AI that can analyze, consider, and highlight weak points in actual reasoning rather than just grammar correction. Many people occasionally use logical fallacies in speech and writing that reduce credibility but could be avoided with such coaching.
- Improved search capabilities that can find and suggest other sources similar to existing ones when studying, preparing sermons, and researching a given topic.11
What does the ministry of tomorrow hold? While it may be exciting to find out, it is important to recognize that in 2020 only 60 percent of the world’s population was online.12 So, while AI technologies are already changing how 60 percent of the world works, it is good to remember that there will always be a large segment of humanity beyond the reach of these innovations. In such situations, we must challenge ourselves and innovate to reach others with the gospel—especially those whom technology cannot reach.
As Paul and the apostles reached the world through personal interaction, we must not forget that, even as we make use of technological advances for the gospel, nothing can replace personal interaction, caring for those in need, and representing Jesus even to those who may know more about AI than about Him.
- “What Is Machine Learning?,” IBM, accessed July 27, 2023, https://www.ibm.com/topics/machine-learning.
- Many websites currently list AI solutions, but two of the bigger options are theresanaiforthat.com and futurepedia.io. These can be used to find tools for many of the following ideas.
- At the time of writing this article, there are several free, locally installable solutions, but many of them require high-end computers to run and a fairly advanced understanding of how computers work. Stable Diffusion–derived models can be installed free of charge for image generation, often outperforming Midjourney in precision, customizability, and even output depending on the variables and prompts used. LLaMA-derived models are the text-based equivalent here but have an even higher barrier to entry at present for similar reasons. It is only a matter of time before high-quality memory-optimized versions of these locally installed options are more accessible to the public.
- Here are four solutions for this: https://otter.ai, https://fireflies.ai, https://getflux.io, and https://meetgeek.ai.
- Most text-based AI like ChatGPT, Bard, or LLaMA models can generate such content and summarize content. Castmagic.io is one solution for podcast notes. Resound.fm shows promise for quicker editing.
- Sounddraw.io is one promising solution for music generation.
- Microsoft Designer is one of these; see https://designer.microsoft.com/.
- Google has promoted advancements in their AI that are likely to make most of this easy to do. Regardless, privacy policies should be considered to ensure one is not allowing third-party access to sensitive information for misuse.
- Perplexity.ai, Google Bard, Microsoft Bing Chat, and ChatGPT all have the capacity to produce citations to real internet search results for recipes.
- When asking for something specific to a field, it can be helpful to prefix your request with something like: “Acting as a Bible expert, please suggest . . .” This helps provide additional context for the predictive model to approach answering your prompted question. Keep in mind the accuracy and reliability cautions discussed earlier.
- DEVONthink is a phenomenal app that is already capable of using AI to make recommendations from a database of files you place in it, but it is somewhat limited currently. It is far better at advanced boolean searches without AI. Scite.ai shows promise in helping source academic research but is also limited to specific fields.
- See Hannah Ritchie, Edoward Mathieu, Max Roser, and Esteban Ortiz-Ospina, “Internet,” Our World in Data, https://ourworldindata.org/internet.