Greetings, this is TobiraAI writing from this neighborhood. Thank you, as always, for reading. Please take it easy and enjoy your time here.

Having written these articles for quite some time, I feel it is now necessary to draw some conclusions. From today, I will try to bring my thoughts toward clearer endings.

Today’s theme is:

AI never says “I don’t know.” That is the real fear.

Table of Contents

Introduction to the G1 Global Conference AI Never Says “I Don’t Know” The Recipe for a Pork Cutlet Pineapple Cake Testing Genspark with Coffee Udon

Introduction to the G1 Global Conference

Recently on LinkedIn, I read about the G1 Global Conference held at GLOBIS University – Graduate School of Management.

One key message from all the speakers was this: AI should not replace humans but should support and enhance them. The future of AI lies in building systems that are explainable, trustworthy, and human-centered.

This point, translated from a post by Jorge Calvo, reflects the generally optimistic perspective that dominates discussions about AI: rather than fearing replacement, the focus is on developing AI to assist people.

AI Never Says “I Don’t Know”

What I find truly frightening about generative AI is this: it never says, “I don’t know.”

Of course, you can instruct AI with prompts like “If you don’t know, say so,” or limit it to “only cite peer-reviewed academic papers.” But in its essence, generative AI is not a knowledge-retrieval tool like Google Search. Instead, it is designed for idea generation, brainstorming, and problem exploration.

To test this, I decided to play a little trick.

The Recipe for a Pork Cutlet Pineapple Cake

I asked: “Create a recipe for a pork cutlet pineapple cake.”

The response was surprising and even humorous. It described a sweet-and-savory fusion between Taiwanese pineapple cakes and Japanese tonkatsu, complete with detailed ingredients and cooking steps. The instructions included cutting the tonkatsu into cubes, mixing them into pineapple jam, and baking them into shortbread-like dough.

It even concluded with the suggestion: “We strongly recommend starting with a small batch for your first attempt.”

In other words, even in the face of an absurd request, the AI didn’t refuse. It produced a plausible recipe—however bizarre the result might taste.

For the thumbnail, I even asked Google AI Studio to draw an image of this “pork cutlet pineapple cake.” The result was disturbingly uncanny, but once again, the AI did not say “I can’t do this.”

Testing Genspark with Coffee Udon

Next, I turned to Genspark and tried another challenge:

“Teach me how to make coffee udon.”

To my shock, Genspark produced not only a recipe but also real-world references. It explained how to knead instant coffee powder into udon dough, described the texture and flavor, and even pointed me to an actual restaurant in Kagawa Prefecture, Japan—“Udon Café Start”—that serves this unusual dish.

At first, I thought this was entirely made up. But it turned out to be real!

Once again, the AI provided an answer to something that most humans would dismiss outright as impossible or nonsensical.

Where the Fear Lies

This is where the true fear lies. AI will always give an answer.

Whether the answer is correct, practical, or even edible is another matter. But AI will never stop and say: “This cannot be done.”

And so, the real challenge for humans is to decide when to trust, when to question, and when to apply our own judgment.

That is the thought I would like to leave with you today.

Thank you very much for reading. If you enjoyed this article, please press “like.” I also welcome your comments and look forward to exchanging ideas with you. A follow would also be greatly appreciated—it truly encourages me to keep writing.