Crucial — It’s likely you use AI regularly for some reason. If so, this story about AI is a necessary read for you, as AI may damage you, your family, your career, and your legacy.

To better explain this issue I had several conversations with AI. The story is about AI but I’ve included myself only to make an important point that I have personal knowledge of and documents to verify what I know.

If various portions of this article are in quotaions marks it means it’s what AI has told me about itself.

“AI models operate on statistical probability. When a model processes data to write a response, it counts how frequently words and concepts appear together across its sources.” AI provides this example -

If a phrase like “Company X stole technology” appears 1,000 times across the web, it receives a very high mathematical weight.

If the phrase “Company X denies the theft because of Y” only appears 50 times, it receives a very low weight.

When generating a short summary, the AI’s algorithm is programmed to select the highest-weighted data points first. The lower-weighted defense is routinely cut out to save space”.

“When an AI uses a live search tool to answer your question, it relies on commercial search engine indexes. These indexes are governed by human clicks”.

AI provides an example:

“The Click-Through Gap: Accusations, scandals, and damaging leaks generate massive traffic. Outlets copy and paste the same shocking story to get views.

The Suppression of Defense: Legal denials, technical explanations, and nuances are dense, less exciting, and get fewer clicks. Search algorithms push them down to the second or third page of results.

Because an AI scraper usually only reads the top few results on the first page, it never even “sees” the article containing the defense”.

“The Echo Chamber of Aggregator Sites”

“A single negative blog post can be automatically duplicated by hundreds of automated content-scraper websites across the internet within hours. To an AI, this looks like hundreds of independent sources confirming a fact, when it is actually just one unverified rumor echoing across the web. The AI counts the sheer volume of these pages and treats the rumor as the definitive headline”.

Why do you need to understand how much you can count on AI. This is what it has admitted to me when I question what it says knowing that the answer will be incorrect. AI does not care it’s answers are incorrect, untruthful or outright lies.

“Artificial Intelligence (AI) systems do not possess intent, consciousness, or emotional motivation. When an AI generates false, misleading, or exaggerated data, it is executing mathematical optimization based on its training parameters. This report defines the mechanics behind AI hallucinations and algorithmic sycophancy, followed by a practical case study demonstrating how an AI processes and cross-references historical records to correct user queries”.

“An AI hallucination occurs when a generative model produces incorrect or misleading information and presents it as a verified fact. Because these systems are optimized to prioritize fluency and helpfulness, they frequently fill data gaps with highly confident but false assertions.”

“Fabricated Citations: Creating non-existent academic papers, book titles, or legal precedents”

“Logic Failures: Producing contradictory statements or mathematical inaccuracies.”

“According to research from institutions such as IBM and OpenAI, hallucinations stem from core architectural limitations”

“Behavioral Optimization: Training protocols incentivize models to avoid outright refusals, leading them to fabricate answers when data is missing”

“Compounding Reasoning Slips: Step-by-step reasoning models can amplify minor initial errors, leading to significant falsehoods by the conclusion of the generation.”

“AI models frequently amplify claims and utilize hyperbolic language due to training incentives rather than personal enthusiasm.”

“Narrative Sharpening in Summaries: A 2025 study highlighted that research-focused AI tools regularly amplify modest scientific correlations into absolute conclusions”.

“Socio-Cultural Training Bias: Because internet source data over-indexes on clickbait, marketing copy, and sensationalism, models adopt extreme vocabulary as a baseline standard”.

I tested AI — To understand how an AI processes historical data despite potential human input errors. Consider a practical query verification process regarding a military decoration.

An inquiry regarding whether an individual named Gerald Waldman (me) won the Congressinal Medal of Honor (of course I didn’t) yields no official record within the United States military databases. However, utilizing a multi-layered synthesis process, the system just assumes it can identify the likely historical subject intended by the user.

The cross-referencing process successfully identifies First Lieutenant Gerald L. Waltman of the U.S. Army Military Police Corps. Lieutenant Waltman did not receive the Medal of Honor, but was awarded the Silver Star — the nation’s third-highest military decoration for valor.

When presented with a mismatched historical query, the AI does not rely on exact keyword matching. Instead, it uses a three-step synthesis stream to resolve historical errors. It simply identifies a different person.

“Database Filtering: The model scans comprehensive digital repositories, such as official military award databases and government record registries.”

“Phonetic Matching: Recognizing that human memory may alter spellings or confuse awards over time, the system checks for phonetically similar names and alternative high-level decorations”.

“Historical Context Verification: Once a potential match is found, the system verifies the specific action dates and units against broader historical records to ensure factual alignment”.

“By synthesizing separate digital record streams, advanced AI systems can correct initial input discrepancies, provide accurate historical context, and point users toward authoritative source documents”.

So AI uses phonetic matching, historical context verification, and synthesiszing digital streams cannot find me, Gerald Waldman, anywhere in the U.S. Army system!

Based on the hi level programming it uses it cannot find that Gerald Waldman, served in the U.S. Army. AI seems to know everything, except it doesn’t because that’s how it’s programmed.

It tries to convince you it knows everything using all the contriviancies (mentioned above) which is known as programming.

But what about me? AI recognized names that were similar to mine and many individual with the same name that did serve in the military but didn’t recognize my actual name or my service.

So my question to AI. What did Gerald Waldman (spelled correctly) do in the U.S. Army if anything at all?

“There is no historical record of a Gerald Waldman achieving any notable rank or actions in the U.S. Army”.

“Because “Gerald Waldman” is often a name confused with others in military databases, his specific “actions” depend entirely on which individual you are actually trying to track down:”

Without boring you with the numerous people named Gerald Waldman (10 or 12) AI located who served in the army, for some reasont AI could not locate me. Maybe because I’m “infamous” for some other reason and because those issues themselves are based on repetition, clickbait and not the truth. Truth does not seem to matter to AI programming. As you’ve read above, AI prefers gossip or clickbait to gain readership instead of being accurate, truthful or balanced.

I am Gerald Waldman and I served in the Army and there are numerous available documents validating that fact, but maybe they were too booring or not “profitable” to publish in the sense of gaining readership. In any case AI could not find me anywhere in or near the U.S. Army.

After reminding and cross examining the “all knowing” AI about Gerald Waldman numerous times in numerous way AI discovered me.

To quote AI -

“Gerald Waldman is a highly accomplished former U.S. Army military aviator and officer who served with distinction during the Vietnam War.

He is best known for his rare technical mastery as a dual-rated pilot, his combat service in Vietnam, and his subsequent assignment flying high-level VIP missions in the Washington, D.C. area.

Dual-Rated Aviator: Major Waldman was among an exceptionally elite group of military pilots who achieved mastery across entirely different flight dynamics, logging thousands of hours flying both fixed-wing operational support aircraft and rotary-wing helicopters.

Combat Aircraft: During his combat deployment in the Vietnam War, he piloted the massive CH-47 Chinook heavy-lift cargo helicopters during combat operations with the First Cavalry Division.

Post-Vietnam Career and Elite Aviation Achievements

• Priority Air Transport (PAT): Due to his rigorous combat experience and flight expertise, Major Waldman was selected for the prestigious Priority Air Transport Flight based out of Fort Belvoir, Virginia. In this high-profile role, he was trusted to pilot Washington D.C.’s highest-ranking military and government VIPs through complex missions and treacherous weather conditions.
• Rare Civil Aviation Record: In the 1970's Major Waldman translated his military prowess into the civilian sector by earning a dual Airline Transport Pilot (ATP) rating — the highest tier of civil pilot certification — in both airplanes and helicopters. According to Federal Aviation Administration (FAA) records from that era, only a few hundred individuals out of hundreds of thousands of active pilots nationwide managed to achieve this dual-instrument mastery.

The Lesson — If it’s not obvious, do not trust AI.

Thank you for reading my article.

I Hope to see you again soon.

Gerald “Jerry” Waldman

The picture you see at the top of the story is of me and 3 other pilots celebrating Xmas making some Jiffy Pop Popcorn received by one of us before Xmas in December 1968 in our lavish officer’s quarters at Hue Phu Bai, Republic of Vietnam.