Bible Bombshells: Uncovering Eye-Opening Facts Behind the KJV and Septuagint, Part 1

A late night bible study goes down a rabbit hole.... and a discussion with AI about scribal edits ensues.

Translating into Greek to create the Septuagint. AI’s interpretation.

I’ve been reading scriptures tonight (Feb 16, 2025, 11:45 PM - part of my Exodus 90 daily Holy Hour discipline), and I’ve stumbled on something I didn’t expect. I have been reading Daniel in the OSB (orthodox study bible), and realized that the LDS KJV omits all of chapter 3:25 -90. I noticed this difference because I taught LDS Sunday school for quite some time, and the OSB has in Daniel 3:25-90 parts that I didn’t recognize as ever having seen before, so I started to compare. This article will give you a sense of how strong my ADHD rabbit hole tendencies are lol, but that also means it’s long, so a TLDR follows.

Important Definitions and Acronyms

This article uses many acronyms to refer to different scriptural texts. Understanding a bit about those texts and how we refer to them will be important background information for this and a following article.

• Texts and Acronyms:

  • KJV (King James Version - 1769 edition) is the version of the bible used by the LDS church. It has been a seminal and enduring force in American Protestantism, shaping theology, worship, and cultural identity from the colonial era through modern times with its distinctive language, literary quality, and deep-rooted tradition. Created in the early 17th century by King James 1st for the Church of England, its Old Testament was translated from the Hebrew Masoretic Text (See below). The New Testament portion was translated from Greek manuscripts compiled during the Renaissance (called the Textus Receptus). It was created as a primarily Protestant Bible but has also been used by some Catholics. One would think that perhaps there are other or newer Bible versions that would be more accurate and could be used by LDS instead. Still, a quick check shows that the same protestant edits made to the KJV also exist in other Bibles produced by protestant groups in America, like the CBS (Christian Study Bible- including the Ancient Faith edition).

  • NKJV (New King James Version 1987) - a modern English translation of the KJV that updates the archaic language of the traditional KJV while preserving its underlying textual tradition, serving as a crucial bridge for contemporary biblical textual analysis and study.

  • LXX (Septuagint) is the ancient Greek translation of the Hebrew Bible from the 3rd to 2nd centuries BCE. It was created to provide a Greek translation of the Hebrew scriptures, enabling Hellenistic Jews (Jews who were culturally Greek - predominantly Diaspora communities in Egypt) to access and maintain their religious traditions in their common language. The Orthodox and Latin Catholic churches still use the LXX. It is vital to biblical textual analysis for offering insights into early Jewish textual traditions and influencing early Christian scriptural interpretation. It is the version of the Old Testament used by early Christians and is centuries older than the MT.

  • DSS (Dead Sea Scrolls) - a collection of ancient Jewish manuscripts found in caves near the Dead Sea, dating from the 3rd century BCE to the 1st century CE, and are crucial to biblical textual analysis because they contain some of the earliest known copies of biblical texts, offering insights into the evolution and transmission of the Hebrew Bible.

  • MT (Masoretic Text) - The authoritative medieval Hebrew version of the Jewish Bible, meticulously compiled and maintained by Jewish scribes (the Masoretes) from the 7th to 10th centuries CE, and it is pivotal to biblical textual analysis because it forms the basis for most modern Protestant translations while reflecting centuries of textual tradition and standardization.

  • OSB (Orthodox Study Bible) - A standard version of the Bible used by Eastern Orthodox, particularly in English-speaking countries. The OSB uses the Septuagint for the Old Testament and the NKJV (New King James Version) translation of the New Testament.

  • OT (Old Testament) - The first section of the Christian Bible, the OT is comprised of sacred Hebrew/Jewish scriptures that chronicle the history, laws, prophecies, and religious poetry of the Jewish people before the coming of Christ. This compilation of ancient scriptures documents the world's creation, the Hebrew people's history and laws, and their covenant relationship with God. The OT was primarily written in Hebrew and Aramaic.

  • NT (New Testament) - The New Testament is the second section of the Christian Bible, comprising writings that recount the life, teachings, death, and resurrection of Jesus Christ and the establishment of the early Christian (Catholic) Church. The NT was primarily written in Koine Greek with some small parts in Aramaic. Contrary to popular misconceptions, the New Testament was not written in Hebrew.

TL;DR Summary/Overview and Context

As it was late at night and I was sitting in bed, I proceeded to ask questions/find answers using Chat-GPT as my preferred search engine (Instead of Google. A discussion about the uses of AI follows below. As a result, this article proceeds as a series of Prompts/Questions from me to the AI, with answers from the AI. To make it easier to read, I’m putting my questions in as section Headings labeled with “Question to the AI:”, after which the AI response follows, formatted in the substack blockquote style. In general, I proceed as follows:

• I start with a discussion on the AI used, explain the models I used and then discuss why I used an AI and what the benefits are over the “old” method of pouring over books and diving into internet article via Google.

• An investigation into how and why the protestant KJV OT omits Daniel 3:25-90 compared to the Orthodox OSB OT.

• I compare the omitted texts and textual variants and talk about differences with the KVJ as a high level look at the translation variations between the Septuagint (LXX), the Dead Sea Scrolls (DSS), and the Masoretic Text (MT).

• Briefly examine reasons why there are variations, specifically variations or omissions created by Jews in the MT and why. Since the LXX was translated hundreds of years earlier than the formation of the MT (Masoretic Text), the LXX was based on earlier versions of the Jewish texts and was the Greek text used by early Christians. Thus, I assume it to be more reliable and authentic from a Christian perspective because it was not subject to scribal anti-Christian tampering.

• I also have a look at the “Additions to Daniel” texts that the reformers considered apocryphal, and look at where they may have come from and why they are in our scriptures.

• I have the AI look at the significant variants and then focus on edits and omissions that specifically deal with messianic passages that early Christians would have used to support their message.

Can we trust the information ChatGTP is providing to us? Embracing AI for Religious Writing and Research: Why?

I lack formal training in biblical textual analysis, Greek, or Hebrew—and given that it was already late at night (and I was already in bed) —I used ChatGPT-4o to find answers to my questions. For reference, GPT-4o is the current default model for the Chat GTP tool created by OpenAI.

About AI LLM’s and Models

Many AI-powered large language models (LLMs), like ChatGPT, allow users to select from a variety of different models within the tool. These LLMs are trained on vast amounts of publicly available (and some non-public, copyrighted) data from the internet. This includes both pro-LDS and anti-LDS materials and more neutral perspectives—though all sources inevitably carry some bias. This broad dataset allows AI to present arguments and insights from multiple sides of an issue without being filtered through any single biased source (including myself).

Different models have distinct capabilities and performance. Changing the model is akin to swapping the engine in a car—in this case, altering the “brain” of the AI. Some models are optimized for speed and efficiency, while others require more computational power but provide deeper reasoning. My initial analysis, before compiling this article, was conducted using the GPT-4o model.

The 4o model represents a significant leap forward in language understanding and generation compared to previous iterations. It excels in advanced language comprehension, contextual awareness, creative and analytical generation, multilingual proficiency, problem-solving, and even has multimodal capabilities (processing both text and images). These capabilities are pretty amazing. After my stroke I had an EEG and put one of the images from the EEG into GTO-4-o and it was able to analyze the entire thing and tell me what was going on with as much detail as I needed. It also told me that it saw indications of ADHD and ASD (both true) and what some of the effects would be (also true.) Currently, GPT-4o is the best general-purpose model available from Open-AI (the maker of Chat GTP.)

At first, I had no intention of publishing my conversation with the AI—I was simply curious. However, after reviewing the discussion, I figured I may as well turn it into an article. That’s the thing with being on Substack, you always need to be looking forward to what you are going to write next. Publishing for an audience also has different considerations than personal inquiry and despite the fact that most of the analysis was done with AI, I’ve spent just as much time assembling this article and making it digestible as any other.

My first consideration was what people’s objections and questions would be regarding the efficacy of the information I received. I determined that GPT-4o has a hallucination rate of 1.5%. In AI terms, “hallucination” refers to fabricated or inaccurate information. While 1.5% is significantly lower than the first versions of Chat-GPT (which some estimates of hallucination rate were between 20-30%,) it still means the output may not be 100% reliable, but what is?

Intending to turn this into an article, I wanted to verify the accuracy of the AI-generated responses. However, I do not have the time, attention, or academic expertise to fact-check everything manually. I’m not a trained textual analyst, nor do I hold a Ph.D. in Koine Greek (the Greek of the Septuagint and New Testament) or Hebrew, making it impossible to assess translations against source texts manually.

To check accuracy, I ran the existing conversation through a newer AI model with advanced reasoning abilities, the o3-mini-high model. The “high” name indicates enhanced analytical capabilities, making it particularly well-suited for verifying complex information. This model reportedly has a hallucination rate of just 0.8%¹, making it the most precise Chat-GTP model available.

The GPT-o3-mini-high model ² is an optimized version that enhances reasoning and data analysis. Some users have noted its advantages³ in scientific research, algorithm development, advanced mathematics, and structured data processing. While it is slightly slower than GPT-4o, it provides greater accuracy by engaging in deeper, multi-step reasoning.

Although the hallucination rate difference between the two models is only 0.7%, using o3-mini-high to fact-check GPT-4o’s output would be far easier and much more feasible than attempting to verify manually (I wouldn’t feel comfortable doing the manual activity and trying to pass off the results as accurate or reliable.) I fed the entire conversation into the o3-mini-high model and asked it to check the accuracy of GPT-4o’s responses. It confirmed the original information but provided additional nuance and clarification in certain areas.

As a writer interested in religious and biblical studies, I approach Artificial Intelligence (AI) with curiosity and caution. I understand why some people are skeptical about using AI, particularly given all of the controversy (and misinformation) about it in the news, with Actors, Artists, and Writers up in arms. I have a background in Technology and studied an older type of AI in college, so I was not afraid to jump in and try it. My experience has shown me that AI, especially Large Language Models (LLMs), can be powerful research tools. Below, I’ll explain how AI assists me in gathering information, reasoning through complex questions, and uncovering insights with greater objectivity.

Understanding AI and LLMs

Before diving into the benefits of using Chat-GPT or an AI to do this research, let me clarify what I mean by AI and LLMs. Large Language Models (LLMs) are AI trained on vast amounts of text. In simple terms, an LLM is like a very well-read assistant with a photographic memory that never forgets, can understand questions posed in natural conversational English, and generates human-like responses. These models have learned from diverse sources – books, articles, historical documents, and more – to predict and form coherent answers. When I ask an LLM a question about a biblical passage or historical context, it uses its broad training to provide relevant information. Essentially, the AI acts as a research partner with a vast library of knowledge.

Unlike traditional computer programs that follow strict rules, modern AI can use logic to reason through problems. For example, the AI can compare different interpretations, summarize arguments, and point out logical connections instead of just fetching a specific verse or commentary. This ability to analyze and synthesize information makes it less like a search engine and more like a collaborator. Recent AI models don’t just parrot facts – they can actually “think” through questions by breaking down problems step by step and adjusting their approach if needed. Knowing this, I used AI to augment my research process and understanding, not to replace it. You’ll see me interject or provide my conclusions and thoughts throughout this article and the next.

Isn’t AI bad? Research Benefits

Because many reading this may not be well versed in modern AI and may be skeptical or even hostile to AI and its uses, I’d like to present what I see as some of the benefits of using this kind of tool (which is becoming ubiquitous and will soon be impossible to avoid.)

The noise you may have heard about AI coming from Journalists, Artists, and Writers are objections to having the LLM AI’s trained on publicly available news, art, and literature. First, news organizations and art/photography websites, chronically underfunded and facing declining revenues, want to be compensated for the work they may have paid for that has been used to train the AIs. Secondly, there is fear that AIs will replace the need for journalists, writers, and artists as they improve. People fear losing their jobs or income from freelance work (both valid concerns). AI can already produce fantastic works of art, but thus far, it’s not very good at creating art that follows your directions nor producing engaging, interesting, and readable literature. That will improve in time. AI could also be used for nefarious purposes (Fraud, blackmail, disinformation, media manipulation, etc.), so research is underway to create tools to combat this. Not like our media hasn’t been manipulated all along anyway, but now computers can do it faster and more effectively and spotting the fraud is getting very difficult.

• Natural Language Understanding: An AI can understand the nature of my inquiries and anticipate my wants and needs, which I can communicate in plain English. In contrast, Google or other search engines generally don’t understand the questions you may be asking them. Instead, search engines do sophisticated pattern matching on the text you enter into the search bar. There is a massive difference between the two, which is why OpenAI has released its own AI-based search engine and Google is racing to mature and integrate its own AI into its search engine product. The AI industry is hot right now; billions are being invested, the rate of progress is lightning fast, and it’s only going to accelerate.

• Access to Diverse Sources: AI can quickly gather and structure information from many sources. Instead of manually combing through countless books and articles, I can ask the AI, and it will find relevant information from different perspectives and scholars. I’m less likely to miss important data because it wasn’t on my bookshelf or was buried on page 2 (or deeper) of a Google search result where I would never find it.

• Efficient Reasoning and Analysis: Recent AI assistants can reason through complex data or conflicting viewpoints and provide nuanced answers quickly, often accomplishing in seconds or minutes what it would take an educated person hours or days to accomplish. I can also ask the AI to summarize interpretations and look for nuances and bias. It speeds up the analysis process because it can handle multiple tasks simultaneously (like comparing texts or checking cross-references). It is important to note that AI doesn’t just give shallow answers; it can break down problems logically. Modern reasoning-enabled AI systems actually “think” through problems by analyzing them step by step and self-correcting mistakes or asking clarifying questions, which is incredibly helpful when grappling with complex subject matter.

• Reduced Bias: One concern we might have is that because an AI may have been trained on biased information, it may inject bias into its answers. However, I’ve found that, in general, a well-trained AI helps reduce personal bias by providing nuance or correcting erroneous assumptions. It presents information from many viewpoints, forcing one to weigh evidence more fairly. By drawing on data from many voices (different theologians, historians, cultures, etc.), the AI offers a composite view that isn’t driven by a single ideology or agenda. In practice, I get a broader, more balanced understanding of a topic. Of course, I still have to use my judgment, but the AI serves as a check against my preconceived notions or implicit biases.

• Deeper Context and New Insights: Because an AI can process so much information, it often uncovers context I am likely to miss. For example, if I’m looking at a Bible verse, the AI might remind me of the historical context, linguistic nuances, or related passages elsewhere in scripture or ancient literature. This has occasionally led to new insights. In one research study on biblical interpretation, scholars found that AI has the potential to significantly enhance biblical hermeneutics by generating new insight and providing rich contextual analysis. I’ve seen this firsthand: AI has pointed out links between texts and interpretations I hadn’t noticed before, enriching my understanding.

Addressing Concerns and Ensuring Balance

I know that using AI in religious research may sound unsettling. Interpreting scripture is, after all, a profoundly spiritual endeavor. I do not hand over the role of interpreting meaning to a machine. Instead, AI is used to augment research and make it more efficient. Instead of spending hours pouring over books, making notes, and using Google to chase down online information and then correlating it together and weaving it into a narrative, AI can do almost all of that quickly and efficiently while understanding the language and complexities of my specific requests in a fraction of the time it would take for me to do it manually. When I make a request, the AI can understand the intentions and needs of my requests and act accordingly, whereas Google can only search for specific words or strings of words. Nobody would have an issue with me manually hunting through books and using Google; this is just an automated and vastly more intelligent way to do that. If the AI provides information or an interpretation, I can always cross-check it with trusted sources (like the Bible, commentaries, or books). AI often points in fruitful directions, but I must discern the truth and draw conclusions on my own, and you’ll see me doing that in the article with asides, commentary, and sometimes snarky comments.

It’s essential to handle AI’s suggestions critically. AI can sometimes make mistakes or present every perspective, including those I know to be less credible. Sometimes, I don’t want an AI to follow a particular line of inquiry or consider specific facts that I know are not relevant or that I believe are suspect. Fortunately, I can ask the AI to do that, and it will follow my guidance. AI also usually cites its sources or at least enables me to find the origin of the information it provides. This transparency is crucial to verify facts and ensure that the insights I’m getting are legitimate.

A Valuable Tool for Deeper Understanding

In my experience, embracing AI technology doesn’t mean compromising rigor or reverence for the subject matter. On the contrary, it allows me to approach issues from multiple perspectives with greater context. The AI’s ability to widen my research scope, provide logical analysis, and deliver balanced, evidence-based insights makes the work of researching and assembling these articles faster and more thorough.

Ultimately, using AI (like LLMs) is about enhancing human effort, not replacing it. I still rely on prayer, reflection, and discipline in my study of scripture, but now I have additional tools that can sift through massive amounts of data, highlight connections, and even suggest interpretations I might not have considered, all in a matter of seconds. For any anyone working with extensive historical and textual data, this capability is incredibly empowering.

I hope this helps demystify AI's role in producing this article. Rather than seeing it as a threat, I see it as a tool that, when used wisely, can lead to more informed, thoughtful, and unbiased explorations of religious topics. My experience shows we don’t have to choose between tradition and technology. With a balanced approach, we can have the best of both: the rich, reflective interpretation from human scholars and the expansive, data-driven support from AI.

<end preamble>

The Beginning of the Rabbit Hole, noticing that the LDS KJV is missing many verses in Daniel.

OK, deep breath. Now that we’ve gotten past the preamble, let’s get on with the meat of the article. The information presented in this short section did not use AI; I noticed this and did the manual work to check against the KJV.

I mentioned at the beginning that I had been reading my scriptures and stumbled on something I didn’t expect. I was reading Daniel in the OSB (orthodox study bible) and came across large sections of the text I had never seen before. Quickly comparing to the LDS KJV, I realized that the LDS KJV omits all of chapter 3:25 -90.

The LDS KJV goes to verse 23 in the narrative, similar to the OSB, but then omits the range mentioned above and picks back up in the KJV at verse 24, which in our bible (OSB) is verse 91.

The omitted verses appear to be liturgical-style texts and songs sung by Shadrach, Mishac, and Abednego while in the fiery furnace. i.e., they are singing liturgical-type songs in the fiery furnace. I recall one of my parish priests saying that the bible of the early church contained part of the church’s prayers and liturgical worship. I had expected this to be only in the New Testament (NT), but I was now reading something in the OSB that appeared to be stylistically liturgical. I haven’t been Orthodox that long, and I haven’t attended an Orthodox Seminary, so I could be mistaken in identifying it as appearing liturgical.

Our verse 91 says “now Nebuchadnezzar heard their singing and marveled and rose up in haste and said to his nobles, “did we not cast three men into the fire?” the KJV used in the LDS church has as its verse

24, “Then Nebuchadnezzar the king was astonished, and rose up in haste, and spake, and said unto his counsellors, Did not we cast three men bound into the midst of the fire”

In the KJV, there is no reason given for why the King is astonished. In the OSB, he’s astonished because he hears them singing. Makes sense to me. If I’d just thrown a bunch of people in a pit of fire to be burned alive, the last thing I’d expect is to hear them singing.

The KJV also omits a good explanation of why/how those standing near the furnace were killed by it (other than implying that it was exceedingly hot and that perhaps they weren’t being cautious.) The OSB verses 46-48 explain the statement in KJV Daniel 3:22.

22 Therefore because the king’s commandment was urgent, and the furnace exceeding hot, the flame of the fire slew those men that took up Shadrach, Meshach, and Abednego. - KJV

46 Now the king’s servants who cast them in did not cease to stoke the furnace with naphtha, pitch, coarse fiber, and brushwood. (Note that the OSB provides and uses the Hebrew names for Shadrach, Meshach, and Abednego.)

47 The flame shot forty-nine cubits above the furnace,

4 8 and it broke out and burned those it found around the furnace of the Chaldeans.

49 But the Angel of the Lord went down into the furnace to join Azariah and his companions, and shook off the fiery flame of the furnace.

50 He made the inside of the furnace to be as though a dew-laden breeze were blowing through it, so the fire did not touch them at all, or cause them pain, or trouble them.

Nelson, Thomas. The Orthodox Study Bible: Ancient Christianity Speaks to Today's World (pp. 1244-1245).