The Autocorrect Endgame
A couple of years ago Google rolled out a feature in Gmail where it'll try to predict and finish your phrases automatically. I'm a programmer, writing isn't my strong suit. I should be happy that they're making it easier for me, right? It's a seemingly innocent feature; a nice quality of life improvement. But it unsettled me. My gut said no.
In a world where many of us utilize autocomplete to help write much of the text we create (I'm even using it to write this now), there's an extremely slippery slope between an editor and a co-writer. Personally, I find myself comfortable when a computer catches my typos or even highlights grammatical mistakes. But when Gmail started getting ahead of me, trying to finish my sentences, it crossed the line.
When autocorrect starts being proactive—instead of reactive—it's planting its flag in your creative process. It is writing words for you to react to.
Silicon valley's tireless pursuit of rounding all the corners off of any friction that personally slights them brought us here, and led us to ChatGPT.
The unease I felt when Google rolled out its text suggestions bubbled up again when a coworker told me they had prompted ChatGPT to write an email to respond to a plumber. With just a quick prompt and some edits, they had a full email ready to send without having to write it themselves1. They had become the autocorrect, ChatGPT the author.
Text generation (Gmail suggestions, ChatGPT, and the like) erodes the individualism that writing can provide, and threatens to homogenize human interaction on the web. What does a computer know about you? About what you want to say? It knows the words you've just written and the word that's most likely to be next. If everyone is pointed towards the path most traveled—towards the most likely text—we end up as a simulacrum of the most average human; regurgitating homogenous phrases devoid of individual style or voice. Just a computer talking to itself.
Maybe I'm overreacting. Maybe this is good. Who really wants to write the same casual-professional boilerplate text for every transactional back-and-forth?
Personally, I'm going to keep using reactive autocorrect. I don't think anyone can bear to live without it anymore. But as someone who isn't good at writing, and is still trying to find their voice, I don't want an AI to whisper words in my ear and sway my style toward the statistical average.
I'm not even going to try to sort out my thoughts on how we write emails professionally out of respect, not necessity. And whatever that means about people feeling comfortable using ChatGPT for a plumber↩