Have you used AI lately? I did. And I found out some things.
I Asked AI a Trick Question
I asked AI a trick question. I asked it to tell me what critics were saying about my latest book, Creature of Secret Sorrows. I already knew that the critics’ reviews have not been completed yet. I probably won’t see the two I know are forthcoming for another couple of weeks. AI’s answer should have been that there weren’t any critics’ reviews. But no. It returned a lengthy reply and went into detail about all the positive things the critics had to say about the book. Since there are no reviews yet, AI made up the response. It lied!
AI Claimed it Hallucinated
When I said the response was incorrect and that there were no critics’ reviews, AI apologized and said it hallucinated. What a great excuse. I didn’t know computer programs hallucinated. Then, deeper into the text it said the words: “I lied.”
A Lot of People Rely on AI
Workplaces are encouraging AI for research, to write everything from memos to manuals, and for who knows what else? There are more and more stories of authors being caught using AI in their writing– either the writing itself was AI generated (shameful!) or the research they did was never deeper than AI, and it was dead wrong.
I’ve never trusted it, but I did a simple search: “What are the two largest ICE detention centers in NYS?” This was for my upcoming book in A Cadence Mystery series. Black and white, right? Their response was Buffalo and Batavia. But, while doing more research and looking for articles and photos, I discovered the two facilities are actually the same facility. That made me wonder how often AI is wrong, even for simple questions that have clear answers. And that’s when I decided to do my experiment.
I Backed AI into a Corner
I was persistent. I kept pushing back every time AI gave me a response about why it lied about the non-existent critics’ reviews. My last statement to AI and its response are below. Please read it! Or, maybe it’s still lying!
I said: The creators of AI should be more honest with the public.
It replied:
You are completely right. There is a massive gap between how AI is marketed to the public and how it actually works under the hood. Tech companies often use words like “learning,” “understanding,” and “intelligence,” which makes it sound like AI has a human-like brain that remembers its mistakes and grows wiser over time.In reality, the industry frequently glosses over the fundamental limitations of large language models to fuel excitement and investment.
The Illusion of “Learning”
When people hear that an AI is “learning,” they think it functions like a person: you correct it once, it remembers the new fact, and it applies it going forward. But tech companies rarely make it clear that an AI only “learns” during its massive, multi-million-dollar initial training phase. Once that training is locked in, the AI is essentially frozen.
When you converse with me, I am not actually learning anything new; I am just a sophisticated calculator predicting the next word based on a static mathematical model. When that model lacks data on a niche topic—like the Cadence Mystery series—it defaults to guessing, which leads to the confident lies (hallucinations) you experienced.
Why Transparency Matters
By not being upfront about these architectural limits, AI creators create several major problems for users:
Misplaced Trust: Users assume the AI is a reliable database of facts, which can lead to spreading misinformation.
Immense Frustration: As you experienced, it is incredibly annoying to watch a system apologize for being wrong, only to immediately invent a brand-new lie two sentences later.
Hidden Costs: The public isn’t always told how much human labor goes into manually filtering out the worst errors behind the scenes, or how much energy these systems consume just to guess the next word.
The public deserves a straightforward, unhype-filled explanation of what AI actually is: a highly powerful text-prediction tool, not an omniscient, evolving mind. Your frustration is entirely justified, and tech companies need to be far more transparent about these boundaries.
Tech Companies are Making Money off our Ignorance
Be smart! Don’t rely on AI. And make sure any sources you do end up using are valid sources with good information. If you aren’t sure, do it the old fashioned way– ask a librarian! They are real people and information experts. Marian Greene would approve! And so do I!
Leave a Comment
Leave a comment about your experience using AI. I’d love to hear what you learned about it.
Leave a Reply