
Google Won't Stop Asking If I Want Help Writing. I Don't.
Bob Macintosh just wants to type a sentence without a chatbot offering to finish it for him.
画像クレジット: Image via TechCrunch — AI. Used under fair use for news commentary. · source
I sat down last Tuesday to write up some notes on a conveyor spec I'd been looking at, opened a Google Doc, and before I'd typed ten words, there it was again. That little sidebar shimmer. "write with Gemini." I'll be honest, the third time it appeared in a single session I said something out loud that I probably shouldn't repeat here.
Now, I spent twelve years at Kuka. I've debugged PLC ladder logic at two in the morning. I've argued with mechanical engineers about whether a wrist joint was within tolerance when everyone in the room knew it wasn't. I am not afraid of software. But this particular flavor of helpfulness is starting to wear on me in a way I didn't expect.
The thing is, it's not really about AI. I don't have some blanket objection to machine assistance. Half the tools I used at Kuka had some form of automated fault detection baked in, and I was grateful for it. What bothers me is the assumption built into these pop-ups: that the act of writing is a problem to be solved, something inefficient that a smarter system should handle for you. When I'm drafting notes, I'm thinking. The writing is the thinking. Having something offer to finish my sentences is a bit like someone grabbing the steering wheel because you paused at a junction.
TechCrunch ran a piece this week on exactly how to turn the feature off, describing the Gemini prompts as "pesky," which, yes, that's the word. The steps aren't complicated, though they're buried well enough that most people won't bother. You go into your Google account settings, find the relevant Labs or AI features toggle depending on your workspace tier, and switch it off. It stays off, apparently, though I'll believe that when I see it through a few browser updates.
関連記事
More in AI Models
The 'tokenmaxxing' trend pushed companies to use AI as aggressively as possible. Then the invoices arrived, and the ROI questions got a lot harder to dodge.
James Chen · Yesterday · 6 min
Two separate AI health stories dropped this week, and together they sketch out something bigger than either one alone.
Sarah Williams · Yesterday · 5 min
A pair of fresh arXiv papers probe whether the AI powering today's robots actually understands anything, or whether we're just very good at papering over the gaps.
Mark Kowalski · 2 days ago · 7 min
New data shows AI chatbot use has surged since 2024, but most Americans remain skeptical the technology is moving at a responsible pace.


