Are you scared, concerned, excited, and/or profoundly moved by AI?

I am all of the above. Yet I use it every day in almost every aspect of my life. However, I am writing this article with my human intelligence—my “HI.” (Although I will admit that I will run this article through Gemini or Claude before I publish, and Grammarly is having its way with me as I type this.)

Lately, almost every conversation I have with friends and new acquaintances ends up on AI.

There was the owner of two holistic, non-toxic nail salons in Brooklyn (with an adjunct wine bar). He uses AI proficiently to generate bank-ordered business plans and to create his Instagram content.

And the environmental lawyer and her husband, a professor. She “hates” AI, is being forced to use it, and is deeply concerned about the damage it is doing to young human minds and the environment. Her academic husband was skeptical, but less concerned. But as I shared my productive use of AI, they warmed up. He even said he would use my personal maxim—“AI is only as good as the HI using it” in his class lectures.

I see it everywhere:

• The 82-year-old professional photographer and teacher who was finally convinced to use ChatGPT by his son and now finds it invaluable for research.
• The young career nanny, looking to expand her career into consulting on the responsible use of social media by children, is concerned about the environmental impact of AI data centers.
• A corporate event producer who has increased her productivity immensely, using AI to plan agendas and visualize physical space plans. Fought it at first, but now she is never going back.

So, back to the point of this article.

AI is smart because of all the human-generated content it relies on. It connects the dots because it has read hundreds of trillions of human words. It synthesizes this chaos into order, allowing us to create art, write more persuasive emails, automate workflows, draft complex RFPs, and write code faster than ever before.

But make no mistake, we are entering a new phase.

AI makers are effectively running out of high-quality human text to feed the machines. To compensate, AI companies are beginning to use “Synthetic Data.”

This creates a risk researchers call the “Ouroboros Effect”—the snake eating its own tail. If AI is trained on text generated by other AIs, the models eventually suffer from “Model Collapse.” The output becomes weird, bland, or hallucinated.

This actually strengthens my argument: AI desperately needs “You” (the human) to keep it grounded. Without fresh human input, the system degrades.

My Experiment: The Proof is in the Negroni

I have many interests, but a significant one is craft cocktails, particularly the iconic Negroni. I have spent years on a quest to discover, craft, and taste over 100 unique variations of this classic drink, documenting the journey on my blog, “100 Negronis.”

I decided to use this personal hobby to test AI’s “smarts.” I asked Gemini: “Is Peter Conway good at making Negroni cocktails?”

Gemini replied, “Yes,” and reported, “I appear to be highly skilled.”

To ensure this wasn’t just AI flattery (or a hallucination), I asked it to explain why it thought that. I demanded an exhaustive brief of the evidence it used to reach that conclusion. Its report was startlingly accurate, citing specific data points from my years of writing without making up a single detail:

1. Depth of Experimentation: It didn’t just see a few recipes; it parsed a decade of distinct entries, recognizing a systematic exploration of the cocktail’s architecture rather than random mixing.
2. Chemical Creativity (The “Riffs”): It identified specific, non-standard flavor profiles I had engineered. It “understood” the culinary logic behind my Smoky Fig (noting the use of caramelized fig syrup and a flamed orange peel). It flagged the Ne-Straw-Groni, recognizing how I used strawberries to mellow the bitter Campari. It even cited the Simply Blood Orange as an example of balancing heavy amaro with fresh citrus.
3. Technical Precision: Most telling was its focus on my technique. It cited my method of “rinsing” a chilled coupe glass with Grappa (a trick I reverse-engineered from the Via Carota Martini) to leave just the aroma without the burn, flagging my notes on dilution control as indicators of a professional-grade skill set.

But here is the catch: Gemini doesn’t have taste buds.

It doesn’t know if my drinks are actually good. It only “knows” I am good because I wrote about it using my human intelligence. It is reflecting my own work back to me. If I hadn’t created that content of over 100 Negroni cocktail recipes and the stories around each one, if I hadn’t put my “HI” into the world, the AI would have nothing to say.

The AI wasn’t smart. It was just quoting me. And I will admit, in a very organized and cogent fashion.

The next time you worry that AI is replacing you, remember: It is starving for your input. Without your “HI,” AI is just an empty echo chamber.