Somewhere in the West, a server room is humming right now. The room is cold in a way that has nothing to do with weather. Somewhere downstream from that room, a river is running a little warmer than it was last summer. Somewhere near that river, a child is brushing her teeth. The water in her cup was clean this morning. By the end of the day, it might not be.
Here, on this Monday, the kitchen is bright in a way I have stopped noticing. The coffee is still unbrewed. A neighbor’s Quran recitation is already drifting through the wall from a small speaker — someone streamed it on Spotify the way most people in this neighborhood do, before the day has asked anyone a question. The motorbike down the street has been idling for a few minutes — a father, probably, warming up before the commute. The week is still in the hands of the coffee, and the coffee is in mine.
I did not know any of this two years ago. I knew AI as the polite box on the side of my screen that waited for me. Something that is there. Something that does not ask me anything.
Then I read a number. Training one large model, in one facility, in one run, evaporated seven hundred thousand liters of clean freshwater. I have been carrying the number around like a stone in my pocket. Seven hundred thousand liters is not a number that argues. It just sits there, heavy and patient, waiting for the next question.
Are we using this wisely yet?
I ask it at different hours, but mostly in the mornings. While the coffee is starting to drip. While typing a message to someone I will probably never meet, asking the machine to make the sentence sound a little softer. While sitting on the floor with a notebook open to an empty page, the AI on my lap waiting, very patiently, for me to type the first word.
The waiting is the part that gets me. The model does not get tired. The river does not get tired either. Somewhere between those two not-tireds, the cost is paid. The cost is paid by a place with a name, by a person whose day starts at five. The cost is paid so that my sentence can sound softer, and so that the coffee can drip, and so that the notebook page can stay empty a little longer.
I have used AI to translate a hundred-page audit from a language I don’t speak well. I have used it to draft a difficult letter to a regulator I was afraid of. For those things, the model earned its keep. I would ask them again.
But there is another kind of prompt. A greeting reformulated. An email polished until the edges disappear. A first draft of a thought I already had. Each one small. Each one a polite request to a system that is not small, in a place that is not close, drawing on a river that is not infinite. I don’t feel the cost when I send them. That is the design.
A few weeks ago I tried a small habit. Before I press send, I wait one second. One second is barely a breath. But the one second has a question in it: did this need a model? Sometimes the answer is yes. Sometimes the answer is honestly, no, I was just lonely at the keyboard. On the days I send fewer prompts, I sleep a little better. I don’t know if the river sleeps better. I hope so.
The coffee is ready now. I am going to pour it, and I am going to drink it, and it will taste like the morning, and the morning will go on. Somewhere in the West, the server room will still be humming. The river will still be running. The child will brush her teeth tomorrow too.
Are we using AI wisely yet?
I keep asking. I am starting to think the asking is the point.
Related reading:
Discover more from Susiloharjo
Subscribe to get the latest posts sent to your email.