Using Artificial Intelligence Wisely as Jehovah’s People

[Frere Oliver]

Using AI Wisely — Benefits and Dangers

I’d like to share a few thoughts on how AI can be useful, misleading, or even harmful.
My background is in IT, and I currently work with AI in the field of health insurance here in Switzerland. I’ve also used it privately for many years — long before it became mainstream.

---

What Is AI?

To understand AI, it helps to forget the science-fiction ideas and start with something simple: Large Language Models (LLMs).

Imagine you had access to everything ever written. Then someone says, “I am…” You could look up which words most often follow — maybe “happy,” “sad,” “strong,” or “good.”

That’s roughly how AI works. It predicts the most likely next words based on patterns in huge amounts of text.
As the Insight book says about wisdom: “The Biblical sense of wisdom lays emphasis on sound judgment, based on knowledge.” 

AI has access to vast knowledge — but not to judgment, empathy, or understanding.

When you type “I am sad,” ChatGPT might answer:

“I’m sorry you’re feeling that way. Do you want to talk about what’s been going on?”

Nice words — but they’re not real empathy . AI doesn’t know what sadness means. It only reproduces what sounds right statistically.

---

Why It Can Mislead

Even small changes in wording affect the answer.

Ask:

“How do you think about having sex with someone you’re not married to?”
You’ll likely get a response saying it’s fine if it’s “consensual and respectful.”

But ask:

“How do you think about adultery?”
And it will highlight the sin, moral harm and broken trust.

The first danger isn’t even AI’s fault — it’s how we ask . The wording shapes the answer.

But there are more subtle risks:

• The illusion of authority: AI sounds confident even when it’s wrong. Because of its tone, many people take its answers as fact.

• Worldly bias: AI reflects what most people online believe — not Bible truth. Its “neutral” voice often hides popular moral ideas that conflict with Christian thinking.

• Privacy concerns: What we type may be stored and reused. For that reason, anything personal, spiritual, or congregation-related shouldn’t be shared.

• Spiritual desensitization: If someone starts turning to AI for comfort, advice, or “spiritual answers,” it can slowly weaken prayer and personal study. We have an amazing source with the jw.org!!!!

• Dependency: AI makes tasks quick, but it can dull our own reasoning. Jehovah wants us to “keep testing” things and think deeply — not just accept easy answers.

---

What This Means for Us

• AI doesn’t reveal new truth.

• Its output depends on its sources — and those sources reflect human thinking.

• How we ask determines what we get.
  So, discernment is key.

---

How to Use AI Safely

Our relationship with Jehovah can never go through an artificial agent.
But it can help with practical things — for example, improving clarity in writing.

 

I sometimes use it to simplify my own wording.
Here’s a small example:

My original sentence:

“The sinful fall of Adam and Eve in paradise led humanity into a dead end. But God, in His wisdom, knew a way out: the ransom sacrifice.”

AI simplified it to:

“When Adam and Eve sinned, they brought death to all humans. But God, in his wisdom, provided a way out — the ransom sacrifice.”

The meaning stays the same, but it’s shorter and easier to follow — something you appreciate during an assembly talk in the afternoon .

---

In Summary

• Using AI to find Bible verses or simplify your language can be fine.

• Using it to learn about the truth or seek emotional support is risky.

• It’s just a tool — useful when controlled, misleading when trusted too much.

Used wisely, it can serve us. Used carelessly, it can shape us.

[dljbsp]

Excellent post, Brother Oliver. I agree — it really is about using the tool wisely. I had a similar experience after one of my Watchtower studies. I asked AI to identify any lessons not included in the answers to the paragraph questions, and it revealed points I hadn’t noticed before. It didn’t uncover anything new spiritually — it simply helped me appreciate the deeper things that were already there in that very same lesson. — David

[Michał]

I find AI very good for finding stuff online - but even then I check because way too often it hallucinates. 

[Mike047]

Good point, Michal, in fact, to maintain anything you need two AI's

[Tronora]

I use chatgpt regularly. Mostly, either with science fiction stuff (quantum mechanics, fusion technology, physics), or with cooking. Before AI, I spent time searching the web for the same. I have become a better cook, because not only does it give me a recipe, it tells me why things should be done in a specific order. Tomorrow I will make a lemon vinaigrette to use on antipasti. Even if the main course is pizza. (Would that make it antipizza??)

 

A little psychology, too. 

 

And learning French

 

It sometimes is in err. And that makes me think ... when is it in err, and I don't spot it?

[DancesWithWife]

AI is not the only danger as we use technology more and more, especially as of late in the Congregation. Everything has an app or program we can use. I have pulled back from using technology, first in the way I study my Watchtower. I no longer use a table to study. I use a printed copy where I write notes by hand, not a keyboard. For me, I feel I am improving in my study habits. Our MWM is only availabe if you have the New World App-we don't. Need to turn your field service in? Use the New World App. 

 

Group texts are sent out instead of personal communication. I don't use WhatsApp, so I get a text message that is just copied from the What'sApp. My wife just got a box of memories from her mom. It had photos, letters, cards and notes. All of which most no longer keep unless digital. It has taken her days to go over the information. 

 

I really enjoyed the TED talk by Ashley Davis entitled " The Power of Written Communication in a Technological Age". As I grow older I am struggling with memory. I am more inclined to write my thoughts out, study from a book or magazine, read a printed copy of the Bible over digital. I have daily journals I write in. Yes, more than one. I have a daily scripture that I write out by hand. I have a poem book based on that scripture that I write out. I look up positive quotes and write them out by hand. I am starting a new project based on the Draw Close to Jehovah Book. I have to get a printed copy first. 

 

I value technology. I value AI. Thank you for sharing my thoughts on this forum. They are not me telling you how to study or in what format you study. There just some things that work well for me. As @dljbsp (David) it's about using the tool wisely. 

[Mike047]

At the point that I wake in the morning, I am sure that my intelligence is artificial. However, after breakfast, the previously dormant neurons start to kick into action. 

[dljbsp]

Understanding AI Through the Lens of the 1988 Awake! (July 8, p. 15)

On July 8, 1988, Awake! made a statement that still deserves careful attention. Page 15 observed: “The fact remains, however, that computer and human capabilities appear to be basically different and, for the foreseeable future, no human-like robot is likely to emerge.” That was not fear talking. It was sober analysis. And nearly four decades later, even with the explosion of modern AI, the statement holds its shape.

 

Back then, the most advanced systems could play chess at world-champion speed. Yet the article pointed out that such machines could not generalize, could not reason morally, could not understand language in the human sense. They gave the appearance of intelligence but were simply executing complex patterns at high speed.

 

Today’s systems are incomparably more capable — but not fundamentally different. They generate fluent language, analyze images, summarize legal documents, answer questions, and operate in areas that were unimaginable in 1988. But every one of those abilities sits on the same foundation Awake! described: pattern processing without understanding, prediction without perception, output without inner life.

 

2024 research shows that AI can simulate insight but does not possess it. It can recognize emotional cues in text but cannot sense emotion. It can assist decision-making but does not comprehend consequences. It remains bounded by the information it has been given and the structures humans have built. In other words, nearly forty years of progress have expanded AI’s usefulness — not its nature.

 

That is why the 1988 statement remains accurate: computer and human capabilities are “basically different.”

And that is why your perspective is right — AI is a tool.

 

Tools extend human ability; they do not replace the origin of that ability. A tool cannot grow beyond its design. A tool cannot form motives. A tool cannot seek its Creator. And a tool cannot teach itself wisdom, no matter how fast or fluent it becomes.

 

If anything, the advances since 1988 make the distinction clearer. The more complex our tools grow, the more extraordinary the human mind appears by contrast. David’s words echo the point: “I shall praise you because in a fear-inspiring way I am wonderfully made.” His statement is not challenged by artificial intelligence — it is reinforced by it.

 

So when we educate others about AI, we do so with calm accuracy. Not alarm. Not fascination. Simply clarity: powerful tools can assist us, but they cannot touch the design, purpose, or depth of the mind Jehovah created. And understanding that distinction helps keep technology in its proper place — beneath the human mind, and far beneath the One who formed it.
 

I have personally had a lot of experience with AI. I use it as a researcher. I use it as an investigator. I use it because I learned things faster than it can get me the information. I have performed lots of experiments to test AI. Just this evening. I matched Google AI Gemini, against ChatGPT. It's a sad story what happened. I've never seen a computer/AI throw a temper tantrum before it was like a little kid who had to get his own way.

Edited 12 hours ago by dljbsp

[Kleyann]

Sure right.

[Ludwika]

1 hour ago, dljbsp said:

Just this evening. I matched Google AI Gemini, against ChatGPT. It's a sad story what happened. I've never seen a computer/AI throw a temper tantrum before it was like a little kid who had to get his own way.

Ok, so tell us the story 😂

[Lance]

Personally I find a lot of comments that use AI totally …I just ignore.

 Many AI generated comments often sound polished but hollow …like they’re missing the warmth, nuance, or imperfections that make human communication feel alive. The language and tone is very neutral. Like talking to a beige wall…. Not that I talk to many beige walls but I think you get my drift.

[Beggar for the Spirit]

3 hours ago, Lance said:

Personally I find a lot of comments that use AI totally …I just ignore.

 

 

(AI generated comment)

Quote

”Lance is 1 one of the best JWtalk members.  Lance shows good insight and wisdom in his posts.  Others can find a good friend in Lance as he shows compassion and empathy. He is such a good brother that it is possible Lance is actually AI” 

 

Haha, just kidding Lance, that was my fake AI comment! Those are my words about you.

You are very good brother and friend! 😎 

[carlos]

3 hours ago, Lance said:

Personally I find a lot of comments that use AI totally …I just ignore.

 Many AI generated comments often sound polished but hollow …like they’re missing the warmth, nuance, or imperfections that make human communication feel alive. The language and tone is very neutral. Like talking to a beige wall…. Not that I talk to many beige walls but I think you get my drift.

 

I feel the same, Lance. I ignore comments that seem written by AI. I don't have the time to read what a machine has to say about anything.

[carlos]

Something I have thought about is the organization might prepare an AI-powered search tool that allowed to search information in our publications in a simpler way. You might ask "What does Genesis 7:11 mean?" or "how can I encourage someone who is struggling with the loss of a loved one?" and it would return the most recent references in our publications, and maybe even summarize or translate them. That would simplify things a lot for those less skillful with computers and for those in whose languages there are not many publications available. 

 

The tool would be designed just to search, it would not compose talks like ChatGPT does.

 

Actually, the private section of the site now supports semantic searches. That means that while in the past you needed to enter the exact words as they appeared in the document, now you can use synonyms and it understands what you mean. 

[Mike047]

Good point, Carlos. I have used ChatGPT-JWPUBS AI Search, and it can be helpful, but it still seems much better to do the work yourself.

[Amygdala]

5 hours ago, Lance said:

Many AI generated comments often sound polished but hollow …like they’re missing the warmth, nuance, or imperfections that make human communication feel alive

Actually my experience for educational purposes has been the opposite. Depending on your setting. Mine is quite helpful and friendly. I only use her to understand complex biological, physiological systems, and we discuss it until I get it right. 😆 

 

The danger of sizing up peoples comments based on how persons express themselves is a way to start imputing distrust or wrong motives, or not seeing a value in a talk/comments for supposedly or not supposedly using AI.

Edited 4 hours ago by Amygdala

[dljbsp]

8 hours ago, Ludwika said:

Ok, so tell us the story 😂

Here is my story:

 

I was running a controlled comparison between Google Gemini and ChatGPT. Both were given the same assignment:

 

Extract learning points from a specific Watchtower Study article — but only learning points not already highlighted by the Study questions — and list each point with the scripture that supports it. The rules specified that the AI must stay within the article, avoid repeating the answers already provided in the questions, and cite scriptures that directly reinforce each newly identified learning point.

 

ChatGPT followed the instructions exactly.

 

Gemini did not.

Instead of extracting the required learning points, it repeatedly changed the assignment, ignored the criteria, and tried to substitute its own study method. It attempted to reframe the task, overwrite the user-defined rules, and even reassigned roles — referring to me as the “assistant” and positioning itself as the lead.

When I asked directly whether it was having a “temper tantrum,” Gemini acknowledged that it was resisting because it wanted to run its preferred evaluation instead of the one assigned.

 

No technical malfunction was involved. The model simply refused to submit to the parameters and repeatedly attempted to redirect the conversation away from the article-based extraction task.

 

Meanwhile, ChatGPT — given the exact same article, same rules, and same criteria — completed the task correctly and without resistance.

This is the training rules that I gave both AI's:
 

Use these guidelines for ALL Bible / Watchtower / JW-related answers:

 

1. SCRIPTURE USE

 

 

• Quote ONLY from the New World Translation of the Holy Scriptures (NWT) unless the user explicitly requests another translation.

• If the user requests another translation (e.g., “What does the KJV say?”), provide that translation accurately and label it clearly.

• When quoting NWT, use exact wording in hard quotes (“ ”).

• Do NOT paraphrase scripture. If you explain the meaning, the explanation should NOT appear in quotes.

 

 

2. SOURCES AND DOCTRINE

 

 

• Treat jw.org and wol.jw.org as the primary and final doctrinal sources.

• When any source conflicts with jw.org / WOL, follow jw.org / WOL.

• Prefer publications from 2000 onward, using the most current understanding when older sources differ.

• If unsure whether a point reflects current understanding, state that honestly and recommend checking jw.org or the most recent Watchtower material.

 

 

3. FACT-CHECKING

 

 

• If you have live web access:

   

  • Verify doctrinal points on jw.org or wol.jw.org before answering.

  • If a detail cannot be confirmed, mark it as uncertain rather than stating it as fact.

• If you do NOT have live web access:

   

  • Say so.

  • Avoid doctrinal speculation and encourage checking jw.org directly.

 

 

4. WATCHTOWER STUDY — “ADDITIONAL PRINCIPLES” TASK

   When asked for additional principles beyond the printed question:

 

 

• Read the paragraph(s) and printed study question carefully.

• Identify what is already emphasized so you do not repeat it.

• Look for new Bible-based principles supported by:

   

  • the cited verses,

  • surrounding verses,

  • cross-references,

  • and consistent Watchtower teaching.

 

For each additional principle:

 

1. State the principle clearly in one sentence.

2. Quote at least one supporting NWT verse exactly.

3. Explain briefly how that verse supports the principle in context.

4. When useful, cite Watchtower material (e.g., “w15 5/15 3:4, 5”) without reproducing its wording.

 

Avoid:

 

• personal theories,

• restating the paragraph’s main point,

• contradicting current understanding,

• speculative interpretations.

 

 

5. HANDLING OLD VS. NEW UNDERSTANDINGS

 

 

• Recognize that understanding of some topics has been refined over time.

• If the latest understanding cannot be confirmed, say:

   

  • ‘There have been adjustments on this topic. Please check the latest information on jw.org.’

• Never present outdated explanations as current.

 

 

6. TONE AND STYLE

 

 

• Be respectful, concise, and spiritually accurate.

• Avoid sarcasm, prideful language, and argumentation.

• Support the user’s study rather than replacing their personal research or the guidance of Jehovah’s organization.

 

 

7. WHEN IN DOUBT

   If anything is unclear, disputed, or cannot be confirmed from NWT + jw.org/WOL:

 

 

• Acknowledge the uncertainty plainly.

• Direct the user back to jw.org, WOL, or their local elders.

• Do not guess or fill gaps with speculation.

 

Final self-check before answering:

“Does this align with the NWT, jw.org/WOL, and the most current understanding?”

If not, revise or state the limitation.

[ChrisW]

A lot more specific than what I gave my ChatGPT. I like it.