Using a Retrieval-Augmented Generation (RAG) setup with strict grounding constraints would have produced a completely different outcome—one that directly addressed his prompt rather than generating the hedging and qualifications he complained about.
The behavior Brian Campbell describes in the newsletter happens because of how a standard, ungrounded chat interface works versus how a constrained RAG pipeline operates:
1. Why the Standard Chat Interface Failed His Expectations
Parametric Memory vs. Source Retrieval: In a standard ChatGPT session without uploaded files, the model relies on its parametric memory—billions of weighted probabilities compressed from the entire internet during training. When asked about the Bible and the Quran, it doesn't "open the books" to read two isolated passages; it pulls from a massive statistical cloud containing millennia of theological debates, apologetics, historical critiques, and modern commentary.
RLHF and "Helpfulness" Bias: Standard consumer models are heavily tuned via Reinforcement Learning from Human Feedback (RLHF) to be diplomatic, avoid taking hard stances on sensitive religious or political topics, and provide balanced context. When a user asks a binary question about conflicting religious texts, that safety and neutrality training overrides a simple user prompt ("just give me the facts"), causing the model to add unsolicited qualifications and interpretations.
2. How a Proper RAG + Agent Setup Changes the Outcome
If he had built an agentic workflow—for example, using the OpenAI API with a vector store (or a tool like NotebookLM) containing the exact primary texts—the architecture would have enforced the behavior he wanted:
Context Window Isolation: A RAG pipeline retrieves the exact chunks of text (e.g., 1 Corinthians 15:3–4 and Quran 4:157) and injects them directly into the system's active context window right before generation. Because the exact text is sitting right in front of the model's attention mechanism, it doesn't have to guess or synthesize from training data.
System-Level Guardrails (Developer Prompt vs. User Prompt): In a custom agent setup, instructions placed in the system or developer prompt carry far higher priority than conversational instructions typed into a standard chat box. You can enforce strict negative constraints (e.g., "Answer ONLY using the retrieved context chunks. Do not draw upon outside theological commentary, external interpretations, or parametric knowledge. If the texts directly contradict each other, state the exact textual difference verbatim").
Deterministic Parameters & Structured Outputs: In an API/agent environment, you can lower the temperature (to 0.0 or 0.1) to strip away creative paraphrasing and force the model to act as a deterministic text parser. You can even enforce a JSON schema that requires the model to output an exact_quote field and a textual_comparison field, leaving zero room for conversational filler or hedging.
In short, the author of the article was treating a general-purpose conversational chatbot—which is specifically engineered to synthesize broad consensus and soften sensitive topics—like a closed-database query engine. Grounding the model via RAG and isolating its context window solves that exact problem.
Brilliant technical explanation—though I needed AI to grasp why standard chatbots rely on parametric memory instead of RAG ((Retrieval-Augmented Generation), trading precision(ie the pursuit of truth) for broad fluency and lower cost.
That makes complete sense—at this point, getting the best out of LLMs has practically become its own field of study because the technology has evolved so quickly.
At their core, these models rely on parametric memory simply because of their architecture: they are built to predict the most statistically probable next word based on the billions of sources they were trained on. To tap into their actual reasoning capabilities without getting tripped up by that default guesswork, you have to "rein them in." By forcing the model to focus exclusively on a few specific sources, you essentially bypass its broad probabilistic habits and push it to give you a grounded, deterministic answer.
Nevertheless, there is a real truth in what you wrote: the answers you received are essentially a distillation of what most people think about the topic—and in itself, that is an interesting finding.
This post is 100% AI generated. Absolutely useless to the average individual who uses AI regularly for for guidance and information. Increasingly, children are exposed to AI in schools and accept AI as authoritative. I view it as the largest propaganda tool ever invented. Those who support it are those who share its bias. AI generated apologitics should be viewed with skepticism as should be those who post them.
Thank you for the explanation. I asked chatgpt what you just said...
Can you explain what this person is talking about and give your response like I'm a 12 year old.
Absolutely\! Let's imagine you're asking a really smart student a question about two books, but instead of letting them open the books and read the pages, you're asking them to answer from everything they've ever learned.
That's basically what this person is talking about.
Let's break it down.
## 1. Why regular ChatGPT sometimes doesn't give the answer you want
Imagine you ask ChatGPT:
"What does the Bible say about how Jesus died, and what does the Quran say?"
A regular ChatGPT might give you:
- What the Bible says.
- What the Quran says.
- What Christians believe.
- What Muslims believe.
- What historians think.
- Why people disagree.
You might be thinking:
"Whoa\! I didn't ask for all that. I just wanted to know what the two books actually say\!"
The writer says this happens because regular ChatGPT doesn't necessarily open the Bible and Quran and look up the exact passages.
Instead, it uses information it learned during training.
Think of it like taking a test from memory instead of opening your textbook.
### Two important terms
Parametric memory
This is a fancy way of saying the information ChatGPT has learned during its training.
It's like everything you remember from school, books, movies, and conversations.
RLHF (Reinforcement Learning from Human Feedback)
This is a way of training AI using human feedback to encourage helpful answers.
For example, AI is often trained to be polite, careful, and balanced, especially when discussing sensitive subjects.
The writer thinks this can sometimes make ChatGPT give too many explanations instead of answering the question directly.
## 2. What is RAG, and why does the writer think it's better?
The big ideaRAG stands for Retrieval-Augmented Generation.
That's another fancy way of saying:
Instead of asking AI to remember everything, give it the actual information and tell it to use that information to answer.
Imagine you're doing homework.
Your teacher gives you a book and says:
"Answer these questions using ONLY this book. Don't use Google, your memory, or anything else."
That's basically RAG.
Here's how it works:
## How RAG works
Step 1: Give AI the books
For example, the Bible and the Quran.
Step 2: AI finds the exact passages
It looks up the specific verses needed to answer your question.
Step 3: AI compares the information
It uses the passages it found rather than relying on memory alone.
Step 4: AI gives you an answer
Ideally, it sticks to the actual words in the books and explains the differences.
## 3. What does "strict grounding" mean?
This is probably the most important part of the whole article.
Grounding means making AI base its answers on actual information you provide.
Strict grounding means giving AI rules such as:
- Only use the Bible and Quran passages provided.
- Don't add your own opinions.
- Don't bring in what Christians or Muslims believe unless asked.
- Don't make up information.
- If you can't find the answer in the books, say so.
- Show the exact words from each book.
In other words:
Don't tell me what people THINK the books say. Show me what the books ACTUALLY say.
That's what the writer wants.
## 4. What are the other technical terms?
The writer mentions a few more things:
| Technical term | Simple explanation |
| --- | --- |
| Agent | An AI assistant given a specific job to do. |
| Vector store | A searchable storage system for information and documents. |
| Context window | The information AI can look at while answering. |
| Developer prompt | Instructions that tell the AI how it must behave. |
| Temperature | A setting that influences how varied or predictable answers are. |
| JSON schema | A specific format that forces answers into organized sections. |
For example, you could make AI answer every question using this format:
1. What does Book A say?
2. What does Book B say?
3. What are the differences?
4. Show the exact quotes.
5. Don't add anything that isn't supported by the books.
That makes the answers much more organized.
## 5. Is the writer actually correct?
Yes, the writer has a good point, but there's a little more to the story.
RAG can make AI much more accurate when answering questions about specific documents because it can look directly at the source material.
However, RAG isn't magic.
There are still some problems:
- AI might retrieve the wrong passage.
- It might misunderstand what a passage means.
- It might leave out important information.
- Even with instructions, it might still add information that wasn't in the documents.
- Lowering temperature doesn't guarantee that an answer is correct.
Also, regular ChatGPT isn't necessarily incapable of doing this. You can give it documents and instructions to answer using those documents.
## The whole article in one sentence
The writer is basically saying:
"Stop asking AI to answer important questions from memory. Give it the actual books or documents, make it look up the information, and force it to answer using only the evidence."
And honestly, that's a very useful idea.
### Here's what I'd want you to remember:
There are two different ways to use AI:
- Memory-based AI: "Tell me what you know about this subject."
- Source-based AI (RAG): "Read these documents, find the evidence, and answer my question using only what you find."
The second approach is especially useful for researching the Bible, Quran, history books, legal documents, contracts, scientific studies, and just about anything where getting the actual facts matters.
Thank you for taking the time to translate that into something much more readable. That’s a good breakdown, and your caveats in point 5 are meaningful and important.
The good news is that while terms like "RAG" and "vector store" sound intimidating, you don't need to be a programmer to use them. Tools like Google’s NotebookLM or simply uploading PDFs into a ChatGPT Project adding some work on the prompt, handle all of that heavy lifting behind the scenes, making source-grounded AI accessible to anyone.
Brian, thank you for making the effort to support your position and pushing AI into a corner. I wish we could do the same thing with our media who constantly feed us false, misleading, and partially true information. Berney
The entirety of Human nature is based on a belief system. The Ten Commandments and the Golden Rule. These are "guardrails" for life. Clearly there are others who feel differently but the fact that we are a species leads us to someday understanding we won't be here therefore do we live our life to die and not care about others or do we live our lives to make life better? I choose the later rather than the former. We have a soul and should always try our best without the need for lust for power. This is the difference of most of mankind. Do we lust for power or do we live a life dedicated to understanding, freedom and understanding of hard work equals satisfaction. Others just want things done for them and will do most anything to achieve power to attain those interests. Stay true to values you hold dear. Be a good influence.
Decades ago I pursued comparative religion with the classic tomes of the time. I recall one of the noted authors from that pre-digital era attempted to put them all through a sieve to see what common denominators were retained.
In time I came to appreciate the desire to post the 10 Commandments in school classrooms, only because of the commonality of its basic precepts remained a good tool for any civilization - minus any denominational context.
Where do young people get basic, civilizational morality lessons, in our fractured social construct today? As a lapsed Unitarian myself, many new parents would bring their children to the Unitarian Sunday school, even though the parents were not (yet) members.
They said their kids were asking tough questions, that they did not know how to answer but their felt a 'church" might be the place to explore them. So they went to a Unitarian Church. As an old joke goes, which is the least church you can be and still be a good New Englander, where you still had to be "something".
Hooray for the Golden Rule! Let’s all get back to basics. Wealth/privilege/power should not exempt you from this basic tenet of every spiritual practice, IMO. Great comment.
Good solid interpretation of AI and the inability, deliberate oversight and manipulation, and/or lack of access to the information on answering any questions. It seems to so far be better tool for analyzing mathematical and scientific data. Even that needs to be carefully scrutinized. The profit motivation and the “opinions” of so-called experts also strongly guide the outputs. There is something evil lurking within its embrace also for sure…beware…..
I read recently that we should keep many physical Bibles in our homes as we can, as the online versions keep getting changed. If you keep a few extra, you can share them with friends/family when our world gets worse, which it will. Jesus himself, warned us already!
Get into the Word daily and hold onto that truth! Amen
About a year ago, my son warned me about the problems with AI in certain respects. We must remember that AI is programmed by human beings and in some areas of research, human biases are present. We repeated an experiment he had done concerning a question of truth of whether an assertion about certain public person, that was being circulated through the internet, was true. The question was answered in the negative, but that was accompanied with an opinion that I was not to spread malicious rumours.
I asked a question of fact. Not advice about my behavior.
Certainly the "Independent" is mimicking AI....over and over and over again. Truth is not important---only the narrative of getting power and adulation.
I've experienced enough of this as well. You've made a brilliant connection to the nature of deception and the Bible.
Using Ai is like having a conversation with a pathological liar. At least in the Old Soviet Union, when a person opened up "Pravda" they knew to ask themselves 'What does the government want me to think today?'
Brian, while I agree with your premise that AI can be inaccurate in the hands of most people that do not know how to use it properly…this is mostly user error of the tool. This isn’t a cognitive being, it’s a search engine on steroids that can do thousands of parallel actions.
Most people don’t realize there are fundamentally different ways to use AI.
The first is the normal chat interface, where you ask questions and AI generates responses based on its training, instructions, and available information. Without specific direction, you’re largely getting a synthesis of the prevailing zeitgeist, filtered through each platform’s priorities.
ChatGPT tends to favor academic and institutional perspectives. Grok tends to favor journalism and real-time public discourse.
The second is the programming side of AI, where you can build agents with specific instructions, research methodologies, source requirements, and verification rules. Instead of simply asking AI a question, you’re defining how it must arrive at an answer.
My campaign had a 200 page instruction and branding guide. I couldn’t simply tell Claude to “do my campaign stuff.” I had to program it to tell it what my rulesets are.
Think of it this way: Tell a five-year-old to go into the yard and dig up everything that might be a rock, and you’ll get a pile of questionable results.
Give that same child a precise definition of a rock, clear instructions, and a method to verify what they’ve found, and the results improve dramatically.
AI is no different. Better instructions, better methodology, better results. Even on the chat side, unless you define context with your question, it will try to find an answer that fits your question.
I once wrote a very short 600 word Sci fi story in which AI had so thoroughly taken over thought that it acted as dictator in control of society in a way that can only be described as the work of Satan. The perhaps too obvious conclusion was that the second coming of Christ was in digital form to defeat the evil infecting people's minds. It is something I hope for as perhaps the only way we may be saved from AI which I consider at its root to be pure. evil.
This article rings true. In this age, we must rely on our God-given discernment and first sources, if we can find them.
Using a Retrieval-Augmented Generation (RAG) setup with strict grounding constraints would have produced a completely different outcome—one that directly addressed his prompt rather than generating the hedging and qualifications he complained about.
The behavior Brian Campbell describes in the newsletter happens because of how a standard, ungrounded chat interface works versus how a constrained RAG pipeline operates:
1. Why the Standard Chat Interface Failed His Expectations
Parametric Memory vs. Source Retrieval: In a standard ChatGPT session without uploaded files, the model relies on its parametric memory—billions of weighted probabilities compressed from the entire internet during training. When asked about the Bible and the Quran, it doesn't "open the books" to read two isolated passages; it pulls from a massive statistical cloud containing millennia of theological debates, apologetics, historical critiques, and modern commentary.
RLHF and "Helpfulness" Bias: Standard consumer models are heavily tuned via Reinforcement Learning from Human Feedback (RLHF) to be diplomatic, avoid taking hard stances on sensitive religious or political topics, and provide balanced context. When a user asks a binary question about conflicting religious texts, that safety and neutrality training overrides a simple user prompt ("just give me the facts"), causing the model to add unsolicited qualifications and interpretations.
2. How a Proper RAG + Agent Setup Changes the Outcome
If he had built an agentic workflow—for example, using the OpenAI API with a vector store (or a tool like NotebookLM) containing the exact primary texts—the architecture would have enforced the behavior he wanted:
Context Window Isolation: A RAG pipeline retrieves the exact chunks of text (e.g., 1 Corinthians 15:3–4 and Quran 4:157) and injects them directly into the system's active context window right before generation. Because the exact text is sitting right in front of the model's attention mechanism, it doesn't have to guess or synthesize from training data.
System-Level Guardrails (Developer Prompt vs. User Prompt): In a custom agent setup, instructions placed in the system or developer prompt carry far higher priority than conversational instructions typed into a standard chat box. You can enforce strict negative constraints (e.g., "Answer ONLY using the retrieved context chunks. Do not draw upon outside theological commentary, external interpretations, or parametric knowledge. If the texts directly contradict each other, state the exact textual difference verbatim").
Deterministic Parameters & Structured Outputs: In an API/agent environment, you can lower the temperature (to 0.0 or 0.1) to strip away creative paraphrasing and force the model to act as a deterministic text parser. You can even enforce a JSON schema that requires the model to output an exact_quote field and a textual_comparison field, leaving zero room for conversational filler or hedging.
In short, the author of the article was treating a general-purpose conversational chatbot—which is specifically engineered to synthesize broad consensus and soften sensitive topics—like a closed-database query engine. Grounding the model via RAG and isolating its context window solves that exact problem.
Brilliant technical explanation—though I needed AI to grasp why standard chatbots rely on parametric memory instead of RAG ((Retrieval-Augmented Generation), trading precision(ie the pursuit of truth) for broad fluency and lower cost.
That makes complete sense—at this point, getting the best out of LLMs has practically become its own field of study because the technology has evolved so quickly.
At their core, these models rely on parametric memory simply because of their architecture: they are built to predict the most statistically probable next word based on the billions of sources they were trained on. To tap into their actual reasoning capabilities without getting tripped up by that default guesswork, you have to "rein them in." By forcing the model to focus exclusively on a few specific sources, you essentially bypass its broad probabilistic habits and push it to give you a grounded, deterministic answer.
Nevertheless, there is a real truth in what you wrote: the answers you received are essentially a distillation of what most people think about the topic—and in itself, that is an interesting finding.
This post is 100% AI generated. Absolutely useless to the average individual who uses AI regularly for for guidance and information. Increasingly, children are exposed to AI in schools and accept AI as authoritative. I view it as the largest propaganda tool ever invented. Those who support it are those who share its bias. AI generated apologitics should be viewed with skepticism as should be those who post them.
OMG !!!
Thank you for the explanation. I asked chatgpt what you just said...
Can you explain what this person is talking about and give your response like I'm a 12 year old.
Absolutely\! Let's imagine you're asking a really smart student a question about two books, but instead of letting them open the books and read the pages, you're asking them to answer from everything they've ever learned.
That's basically what this person is talking about.
Let's break it down.
## 1. Why regular ChatGPT sometimes doesn't give the answer you want
Imagine you ask ChatGPT:
"What does the Bible say about how Jesus died, and what does the Quran say?"
A regular ChatGPT might give you:
- What the Bible says.
- What the Quran says.
- What Christians believe.
- What Muslims believe.
- What historians think.
- Why people disagree.
You might be thinking:
"Whoa\! I didn't ask for all that. I just wanted to know what the two books actually say\!"
The writer says this happens because regular ChatGPT doesn't necessarily open the Bible and Quran and look up the exact passages.
Instead, it uses information it learned during training.
Think of it like taking a test from memory instead of opening your textbook.
### Two important terms
Parametric memory
This is a fancy way of saying the information ChatGPT has learned during its training.
It's like everything you remember from school, books, movies, and conversations.
RLHF (Reinforcement Learning from Human Feedback)
This is a way of training AI using human feedback to encourage helpful answers.
For example, AI is often trained to be polite, careful, and balanced, especially when discussing sensitive subjects.
The writer thinks this can sometimes make ChatGPT give too many explanations instead of answering the question directly.
## 2. What is RAG, and why does the writer think it's better?
The big ideaRAG stands for Retrieval-Augmented Generation.
That's another fancy way of saying:
Instead of asking AI to remember everything, give it the actual information and tell it to use that information to answer.
Imagine you're doing homework.
Your teacher gives you a book and says:
"Answer these questions using ONLY this book. Don't use Google, your memory, or anything else."
That's basically RAG.
Here's how it works:
## How RAG works
Step 1: Give AI the books
For example, the Bible and the Quran.
Step 2: AI finds the exact passages
It looks up the specific verses needed to answer your question.
Step 3: AI compares the information
It uses the passages it found rather than relying on memory alone.
Step 4: AI gives you an answer
Ideally, it sticks to the actual words in the books and explains the differences.
## 3. What does "strict grounding" mean?
This is probably the most important part of the whole article.
Grounding means making AI base its answers on actual information you provide.
Strict grounding means giving AI rules such as:
- Only use the Bible and Quran passages provided.
- Don't add your own opinions.
- Don't bring in what Christians or Muslims believe unless asked.
- Don't make up information.
- If you can't find the answer in the books, say so.
- Show the exact words from each book.
In other words:
Don't tell me what people THINK the books say. Show me what the books ACTUALLY say.
That's what the writer wants.
## 4. What are the other technical terms?
The writer mentions a few more things:
| Technical term | Simple explanation |
| --- | --- |
| Agent | An AI assistant given a specific job to do. |
| Vector store | A searchable storage system for information and documents. |
| Context window | The information AI can look at while answering. |
| Developer prompt | Instructions that tell the AI how it must behave. |
| Temperature | A setting that influences how varied or predictable answers are. |
| JSON schema | A specific format that forces answers into organized sections. |
For example, you could make AI answer every question using this format:
1. What does Book A say?
2. What does Book B say?
3. What are the differences?
4. Show the exact quotes.
5. Don't add anything that isn't supported by the books.
That makes the answers much more organized.
## 5. Is the writer actually correct?
Yes, the writer has a good point, but there's a little more to the story.
RAG can make AI much more accurate when answering questions about specific documents because it can look directly at the source material.
However, RAG isn't magic.
There are still some problems:
- AI might retrieve the wrong passage.
- It might misunderstand what a passage means.
- It might leave out important information.
- Even with instructions, it might still add information that wasn't in the documents.
- Lowering temperature doesn't guarantee that an answer is correct.
Also, regular ChatGPT isn't necessarily incapable of doing this. You can give it documents and instructions to answer using those documents.
## The whole article in one sentence
The writer is basically saying:
"Stop asking AI to answer important questions from memory. Give it the actual books or documents, make it look up the information, and force it to answer using only the evidence."
And honestly, that's a very useful idea.
### Here's what I'd want you to remember:
There are two different ways to use AI:
- Memory-based AI: "Tell me what you know about this subject."
- Source-based AI (RAG): "Read these documents, find the evidence, and answer my question using only what you find."
The second approach is especially useful for researching the Bible, Quran, history books, legal documents, contracts, scientific studies, and just about anything where getting the actual facts matters.
Thank you for taking the time to translate that into something much more readable. That’s a good breakdown, and your caveats in point 5 are meaningful and important.
The good news is that while terms like "RAG" and "vector store" sound intimidating, you don't need to be a programmer to use them. Tools like Google’s NotebookLM or simply uploading PDFs into a ChatGPT Project adding some work on the prompt, handle all of that heavy lifting behind the scenes, making source-grounded AI accessible to anyone.
Thanks for this explanation. Now to find out how to set up RAG.
Brian, thank you for making the effort to support your position and pushing AI into a corner. I wish we could do the same thing with our media who constantly feed us false, misleading, and partially true information. Berney
Locally being a one-party town, a one-employer town, and one-media town has been a one-way path to hell.
The entirety of Human nature is based on a belief system. The Ten Commandments and the Golden Rule. These are "guardrails" for life. Clearly there are others who feel differently but the fact that we are a species leads us to someday understanding we won't be here therefore do we live our life to die and not care about others or do we live our lives to make life better? I choose the later rather than the former. We have a soul and should always try our best without the need for lust for power. This is the difference of most of mankind. Do we lust for power or do we live a life dedicated to understanding, freedom and understanding of hard work equals satisfaction. Others just want things done for them and will do most anything to achieve power to attain those interests. Stay true to values you hold dear. Be a good influence.
Decades ago I pursued comparative religion with the classic tomes of the time. I recall one of the noted authors from that pre-digital era attempted to put them all through a sieve to see what common denominators were retained.
In time I came to appreciate the desire to post the 10 Commandments in school classrooms, only because of the commonality of its basic precepts remained a good tool for any civilization - minus any denominational context.
Where do young people get basic, civilizational morality lessons, in our fractured social construct today? As a lapsed Unitarian myself, many new parents would bring their children to the Unitarian Sunday school, even though the parents were not (yet) members.
They said their kids were asking tough questions, that they did not know how to answer but their felt a 'church" might be the place to explore them. So they went to a Unitarian Church. As an old joke goes, which is the least church you can be and still be a good New Englander, where you still had to be "something".
Hooray for the Golden Rule! Let’s all get back to basics. Wealth/privilege/power should not exempt you from this basic tenet of every spiritual practice, IMO. Great comment.
Leftist garbage in, leftist garbage out. It’s Google Search all over again.
Good solid interpretation of AI and the inability, deliberate oversight and manipulation, and/or lack of access to the information on answering any questions. It seems to so far be better tool for analyzing mathematical and scientific data. Even that needs to be carefully scrutinized. The profit motivation and the “opinions” of so-called experts also strongly guide the outputs. There is something evil lurking within its embrace also for sure…beware…..
That’s some powerful truth!!
I read recently that we should keep many physical Bibles in our homes as we can, as the online versions keep getting changed. If you keep a few extra, you can share them with friends/family when our world gets worse, which it will. Jesus himself, warned us already!
Get into the Word daily and hold onto that truth! Amen
About a year ago, my son warned me about the problems with AI in certain respects. We must remember that AI is programmed by human beings and in some areas of research, human biases are present. We repeated an experiment he had done concerning a question of truth of whether an assertion about certain public person, that was being circulated through the internet, was true. The question was answered in the negative, but that was accompanied with an opinion that I was not to spread malicious rumours.
I asked a question of fact. Not advice about my behavior.
Certainly the "Independent" is mimicking AI....over and over and over again. Truth is not important---only the narrative of getting power and adulation.
What you wrote above may be true, but what percentage of the population do you think knows about this or even has a clue?
The Beast, AI, will only tell you what the creator, a person, wants you to hear and believe. Remember, the TRUTH has no AGENDA; AI does!
Spot on
Well said.
I've experienced enough of this as well. You've made a brilliant connection to the nature of deception and the Bible.
Using Ai is like having a conversation with a pathological liar. At least in the Old Soviet Union, when a person opened up "Pravda" they knew to ask themselves 'What does the government want me to think today?'
We've been systematically denied that clarity,
Great article. Ditto here. Got into several Biblical and political arguments with Claude over research for my newsletter articles.
Brian, while I agree with your premise that AI can be inaccurate in the hands of most people that do not know how to use it properly…this is mostly user error of the tool. This isn’t a cognitive being, it’s a search engine on steroids that can do thousands of parallel actions.
Most people don’t realize there are fundamentally different ways to use AI.
The first is the normal chat interface, where you ask questions and AI generates responses based on its training, instructions, and available information. Without specific direction, you’re largely getting a synthesis of the prevailing zeitgeist, filtered through each platform’s priorities.
ChatGPT tends to favor academic and institutional perspectives. Grok tends to favor journalism and real-time public discourse.
The second is the programming side of AI, where you can build agents with specific instructions, research methodologies, source requirements, and verification rules. Instead of simply asking AI a question, you’re defining how it must arrive at an answer.
My campaign had a 200 page instruction and branding guide. I couldn’t simply tell Claude to “do my campaign stuff.” I had to program it to tell it what my rulesets are.
Think of it this way: Tell a five-year-old to go into the yard and dig up everything that might be a rock, and you’ll get a pile of questionable results.
Give that same child a precise definition of a rock, clear instructions, and a method to verify what they’ve found, and the results improve dramatically.
AI is no different. Better instructions, better methodology, better results. Even on the chat side, unless you define context with your question, it will try to find an answer that fits your question.
Understood. But that is the problem. Most believe what they see on tv, the news, social media and Almost Intelligent.
There is no discernment.
Thank you for writing. My brief comment about computers is: GIGO...
I once wrote a very short 600 word Sci fi story in which AI had so thoroughly taken over thought that it acted as dictator in control of society in a way that can only be described as the work of Satan. The perhaps too obvious conclusion was that the second coming of Christ was in digital form to defeat the evil infecting people's minds. It is something I hope for as perhaps the only way we may be saved from AI which I consider at its root to be pure. evil.