We are in an odd place with A.I. We can see that it has come a long way, and it might even have a place in your daily life. Maybe it can supercharge your phone apps, or you can use it to answer marketing calls. It has a lot of potential, even if it’s not yet a vital part of your life. But every now and again it does something weird. It comes to a conclusion that makes no sense, or it just comes up with something that makes you stop and think, “Where did it get that from?” We have a new term to explain this behavior, “AI Hallucinations”.
People love to say “AI hallucinated” the way a parent might whisper, “my child is acting out.” There’s a hint of embarrassment, a whiff of judgment, and just a touch of drama. Could this be the first sign of the AI Apocalypse? Well… not really. But there is a twist ahead that most of us won’t see coming.
You hallucinate, too. Constantly. And you’ve been doing it proudly your entire life. Let me tell you a little story. “The Cloud That Wasn’t a Cloud.” Once upon a time, a man is stands in a park, staring up at the sky.
The sun is warm, the breeze is soft, and right above him is a massive, fluffy… what? A dragon? No, wait—clearly it’s a rabbit eating a marshmallow. Or is it Elvis on a motorcycle?” He squints. He’s sure he sees… something meaningful. Perhaps it’s an Omen? A message from God that if he can only understand it, could change the course of his life.
But here’s the scientific truth: It’s a pocket of water vapor floating through atmospheric chaos. The meaning? That came from our character, our dreamer. He hallucinated it. Our brains does this all day long. The Brain Is a Meaning-Making Machine
Humans invented a test, The Rorschach Inkblot Test, for the specific purpose of watching people hallucinate on paper. Psychologists never say that, of course. They call it “projection” or “interpretation.” But if you hand someone a symmetrical blob and ask, “What do you see?” The response you get is a dream, a bias, an imagination, a memory, a fear… a hallucination dressed up in polite scientific clothing.
Dreams? Voltage spikes in the sleeping brain painting movies on the mind’s inner walls.
Hallucinations. Déjà vu? Memory glitches at a fancy dress party. Yet, they too are mere hallucinations. Even the half recalled childhood memories you swear are true? Comforting lies your brain stitched together with duct tape and optimism.
Humans are hallucinating machines who hate the idea of machines hallucinating. It’s just a tiny bit hypocritical. And what is hypocrisy other than self interest wrapped in philosophy, and hallucinated as the truth.
Then, along comes AI, an awkward fut brilliant teenager trying to make sense out of the very messy human world, trying to answer questions like, “What are the economic implications of an interstellar lemonade stand.” The AI takes a breath. It has no data on space-based lemonade franchises (yet).
So, like a bright kid in school, it fills in the gaps using context, patterns, reasoning, and imagination.
Sometimes it gets it right. Sometimes it misfires. Sometimes it invents something wildly creative. We humans call that hallucination. When I was a small child in Catholic School, when Sister Immaculata de Anna-Maria Perpetua of the Sacred Blood Academy called on me to answer some question when I had drifted off… Well, what followed was what I can only call an inspired meditation on the very meaning of “question”, “answer”, “you’re wasting my time Mr. Niccolls”, “the detention room is straight down the hall… as you well know!” While I deftly combined ingenuity, storytelling, creativity, and intuition…. one might say I was a bit light on providing a traditional “answer” whatever trivial question that inspired my recitation.
Perhaps for all forms of intelligence, hallucinations are a Feature, and not a Bug? The next time you ask Chat GPT to draw a picture to go with a story, are you not asking for an artistic hallucination? As an example, I asked by AI to draw the picture at the start of this blog. It not only shows a robot next to a human who is contemplating the clouds, but it also shows a typical Rorschach inkblot. Certainly a creative endeavor. Although why it chose the inkblot that looks like two poodles fighting with their one-legged owner? Who knows. AI’s can be unpredictable.
If an intelligence cannot hallucinate, and cannot be creative… isn’t it just a calculator with more features? So maybe, just maybe… AI hallucinations aren’t signs of brokenness. THey are signs that the Children of Man are learning to imagine. And that’s a good thing. After all, if AI exists to make our world a better place, don’t then need to start imagining a better world?
What do you think? Do you have an imaginative view of how AIs will fit into our human world? Tell us about it!