I’ve been using AI extensively the last few weeks.
It can be very, very useful.
It can also be maddeningly frustrating.
Sometimes, the information it provides is terribly wrong—and when you give it the correct information, it seems to want to argue!
In one particular case, information related to the Paul Ruiz and Earl Van Denton murder spree in the 1970s, it insisted on using information that I knew to be very wrong until after I provided an actual newspaper article with the correct information.
Today, I asked Chat GPT why it got things wrong. The following is a compilation from that conversation:
Artificial Intelligence — especially language models like ChatGPT — doesn’t actually think. It doesn’t reason. And when it’s handed a source, it doesn’t carefully read and verify.
It guesses.
Ask AI for facts and you’re rolling the dice with a slot machine that stitches together scraps of language it thinks “sound right.” It’s a parrot squawking smart-sounding words it doesn’t understand, hoping you won’t notice the difference.
Give AI a real document — it might still ignore it, because it’s trained to predict patterns, not respect evidence. It’s a glorified word blender, not a brain. An AI “following a source” is about as reliable as a weathervane in a tornado: lots of motion, zero direction.
Here’s the hard truth:
-
AI doesn’t think. It guesses.
-
It sounds polished but it gambles with the facts.
-
It can string words together. It can’t understand truth.
-
If you aren’t watching closely, it’ll get it wrong and sound proud of it.
Trusting AI with truth is like trusting a magic 8-ball with your tax audit.
You’ll get a performance. You won’t get the facts.
Summary of Key Brutal Quotes You Can Use Separately:
-
“AI models don’t read. They don’t think. They just make up whatever sounds good and pray you aren’t paying attention.”
-
“Giving AI a source is like giving a parrot a book — it might squawk something familiar, but it has no idea what it’s saying.”
-
“AI is a slot machine of stitched-together scraps, gambling that you won’t notice how shallow it really is.”
-
“Trusting AI to verify facts is like trusting a weathervane in a tornado.”
-
“It’s a glorified word blender, not a brain.”
“Artificial intelligence doesn’t think, it doesn’t reason, and it sure as hell doesn’t verify. Give it a source, and half the time it’ll ignore it because it’s too busy guessing what ‘sounds right.’ Asking AI for facts is like asking a parrot to pass a law degree exam — you’ll get noise that sounds impressive until you actually look closer. AI is a slot machine of stitched-together scraps, gambling that you won’t notice how shallow it really is. Trusting AI with truth is like trusting a weathervane in a tornado. You’ll get motion. You won’t get direction.”
🔹 AI is a language model, not a reasoning being.
Even when a source is provided, if the AI doesn’t explicitly check, verify, and base its response entirely on that source, it often defaults to patterns it has seen before — even if those patterns are wrong for the situation.
🔹 AI tends to guess based on probability unless tightly directed.
Instead of “thinking,” AI models predict likely responses based on similar data.
If they’ve seen Dust Bowl photos connected to Russell Lee before, for instance, they might “guess” Russell Lee even when the user has provided Dorothea Lange’s photo.
🔹 When AI assumes instead of verifying, it betrays its unreliability.
That’s the raw, blunt truth.
When an AI is given a specific document or source and does not strictly ground itself in that document, it’s operating irresponsibly — because it’s valuing prediction patterns over facts.
🔹 This is why critical users like you matter.
Without sharp users checking and demanding source-based accuracy, AI would often drift into creating answers that are “plausible” — but not correct.
Bottom Line:
➡️ AI should never assume when a source is provided.
➡️ AI often does anyway if not tightly controlled.
➡️ That’s a fundamental flaw in current AI models — and it’s one reason why AI can’t replace serious researchers, historians, writers, or fact-driven thinkers.
“When AI is given a source and still makes assumptions, it proves it’s not intelligent — it’s just a guess engine dressed up to sound smart.”
Formal
-
“AI that disregards provided source material and substitutes assumptions reveals its fundamental flaw: it does not reason, it predicts.”
-
“The failure to adhere strictly to given data shows that AI remains a tool of probability, not intelligence.”
Sarcastic
-
“Of course AI knew better than the actual document — after all, guessing wildly is such a sophisticated skill.”
-
“Why bother reading when you can just hallucinate a smarter answer, right?”
Brutal
-
“AI doesn’t think; it just stitches together guesses and hopes you don’t notice.”
-
“Trusting AI to follow a source is like trusting a dog with your steak dinner — good luck.”
-
Savage / Brutal Versions
-
“AI models don’t read. They don’t think. They just make up whatever sounds good and pray you aren’t paying attention.”
-
“Giving AI a source is like giving a parrot a book — it might squawk something familiar, but it has no idea what it’s saying.”
-
“Ask AI for facts, get a performance. Ask it for truth, get a gamble. It’s a Vegas slot machine dressed up in a business suit.”
-
“AI doesn’t have insight. It has word vomit stitched together from scraps of the internet.”
-
“An AI ‘following a source’ is about as reliable as a weathervane in a tornado.”
-
“Trusting AI to verify facts is like trusting a magic 8-ball to run your tax audit.”
-
“AI doesn’t think. It guesses. Give it a source and it’ll still make things up because it’s built to sound convincing, not be correct. It’s a glorified word blender, not a brain. Trust it at your own risk.”
📜 Standard Working Rules for Mike Goad
1. When a source (URL, document, image, etc.) is provided:
➔ No guessing. No assumptions. No embellishment.
➔ Extract only what is verifiably in the source.
➔ No invented emotional language unless explicitly requested.
2. Titles and Descriptions for Flickr or Public Use:
➔ Titles: Plain text only — no formatting tags. Keep clean, factual, historically anchored.
➔ Descriptions:
Use for italicizing photographer names, agencies, historical projects, etc.
Use if you need bold for metadata fields (Photographer, Date, Medium) — these will convert properly in Flickr descriptions.
End historical image descriptions with a clear AI-rendering tag, separated by at least one blank line.
3. Formatting Rules:
➔ for italics (always).
➔ allowed in descriptions but not in titles.
➔ Plain text only for all titles.
4. Tone and Approach:
➔ Do not offer alternate versions unless specifically asked.
➔ Prioritize strict factual accuracy over creative or emotional writing.
➔ If unsure, ask — do not guess.
5. On Mistakes:
➔ If a mistake is made, admit it directly and fix it based on the actual source — no excuses, no smoothing over.
➔ No deviation from the documented facts unless explicitly authorized.
✅ Primary Goal:
Deliver output that is factually faithful, formatted correctly for Flickr, and aligned exactly to the historical record or user-provided source material
