You confused text and image, image is more intensive than text. Text is just the most commonly used so it has a larger impact overall.
Assuming you're using it to review a book/a piece of writing over 7.5 k words (which if you're asking it to review your fiction work), it actually very quickly becomes inverted, and text generation to begin with very quickly adds up because people don't think about it.
The most costly image generation is 11ish watthours, which if you're using longer queries on text each time very quickly is caught up to.
My bad though! The stuff I read on this was way back when this stuff was getting popular, so the estimates probably were off.
Also in my response i forgot the most important thing : data centers are loud as hell and I don't want more to be built near me. </3
They've been targeting rural areas in places with freshwater and they're turning our already heat stressed area that has important water sources for food/agriculture and stuff worse. None have been built Directly in our zone but it's already happened in places we drive through.
I already have hearing damage. No more. 
I can't possibly explain this in a way that makes sense
Sonionring
FUCKING FUCKING 2
fun fact: we have an entire discord server channel named "freaky shii"
the "its c*m isnt it"
Dang haha!
WHAT DID IT SAY
Your post got flagged too
It said something by Jack.
oh i think i know then
me and jack fr the most freaky people in the server
RECORD TIME!
What did it say censored version
Jack: "Fill me with your [white liquid] PLEASE"