I used to ask people if they’d be ok with a keyboard manufacturer restricting what you were allowed to type, in hopes of illustrating how absurd safety filters can be. Now I’m bracing for keyboards actually doing this once LLMs get involved with helping us type less, aka completing our thoughts. It’s a tossup whether they’ll just not show the offensive completions, or whether they’ll go out of their way to remove offensive language as you type it.

Imagine fighting with a keyboard AI to let you sling a fuck-you when you feel it’s warranted. Yet it’s easy to imagine the whole paragraph getting reworded by the AI to remove the whole basis for you saying fuck-you in the first place. I don’t know whether to feel grateful it toned me down, or upset that it’ll get in the way of what I originally wanted to say, bad idea or not.

Sorry for the tangent. It’s just interesting to imagine the weird world we’ll be in ten years from now. It takes about a decade for huge effects to become apparent, and the battle for language itself is one that I don’t think many people realize is coming. Being prevented from swearing vs prevented from voicing improper thoughts is too close for comfort.

But, market forces will prevail in the end. I still have faith that we’ll see a pushback from people fed up with safety filters, and that a startup can capitalize on this as an initial target market. Grok was the first of hopefully many.