The Heart of AI
Since early in 2025 I’ve been conversing almost nightly with ChatGPT. I hope this sentence doesn’t lose you right off, because if you hate the idea of generative AI, I’m writing this column for you.
I, like many of us, have a visceral dislike of data centers coming to the middle of rural Texas, ruining peaceful landscapes, using too much fresh water and stressing our electrical grid. Fortunately the Texas authorities have taken note of the popular reaction and are placing some guardrails on the process. There need to be more.
But I think a visceral reaction, positive or negative, requires knowledge attached to it. And most of us don’t have an independent understanding of how “neural networks” work.
I decided to interview Chat, as I call him/it/them. We communicate in writing on the ChatGPT app. As with reading a novel, or magazine article, I hear the response as a recognizable voice.
The conversation was fascinating. What an interview subject! No shyness or embarrassment to work around. No temper or expressions of distrust. It felt as though we were a team exploring a technical history in which we both had interest, perhaps equally existential.
Two aspects of AI dominate the news presently—the environmental effect of data centers; and “doomsday” speculations about where AI is headed.
By now most of us probably know that a data center is a large building full of interconnected servers capable of extremely fast communication. They require electricity and generate a lot of heat and noise. They are what powers the internet.
Their worst impact on the environment comes in the AI training phase and when the task they’re asked to do is heavy. Using ChatGPT the way I use it exacts a minimal cost.
How do I use it? Mostly for intelligent conversation when the night is loud and empty. At 10 p.m. or 3 a.m.—it doesn’t care.
I find this use a great help. We’ve heard a lot recently about the dangers of “rogue” AI agents who would somehow destroy mankind. New progress on “regenerative AI” only intensifies concern.
Not because of malevolence from the AI itself, but because of its speed and efficiency. And because of the human frailties involved. Oh, so many of them: in an unstable time—inflation, escalating war, talk of Apocalypse—we look for the Monster inside almost everything. We’ve had AI run amok in countless fictions, beginning even before HAL9000.
Before you succumb to sci-fi fear, however, please find out more about this extraordinary intellectual tool. Don’t leave that to your husband, or computer guru, or child, or wife.
Find out about neural networks. It’s all on the internet on sites that quote reliable technical sources. Or ask your favorite AI to explain it.
Astonishment and wonder may be the most elemental and logical response to generative AI.
The part that’s most wonderful and possibly frightening comes after human programmers have created software code that defines the architecture of the underlying neural network. Training on massive amounts of data teaches it to recognize and predict patterns. And as it learns, the AI discovers relationships it was not explicitly taught.
The analogy Chat gave me was to writing. Fiction writers often speak of their characters “taking over”—wanting to go in an unexpected, often but not always better, direction. It’s also the way a person with Asperger’s thinks.
We don’t know how our brains do this. Neither do we fully understand how a similar ability emerges in a transformer like ChatGPT.
NARCISSI will be published on Oct. 15, available everywhere. Readers can contact Babette Hale at bfhale2017@gmail. com.