Since about two weeks, I‘m noticing that autocorrect on my Iphone is increasingly changing the most common and likely words in a sentence to the most absurd nonsense trash. Sometimes, it’s not even an existing word. I‘m using German and English keyboard on an Iphone 8 and I‘ve never had issues with autocorrect until recently.
Someone noticing similar issues? Did they implement some kind of new AI at some point that just doesn’t understand shit? Unfortunately, I didn’t save or remember any examples of stupid things it corrected.
I’ve definitely noticed this over the last month or so. It’s so bad it changed could to cock and I sent it to a potential client! That was an awkward follow up!
every time I want to type “character” it takes 20 attempts to stop writing “charcuterie”. I’ve never used that word once in my life and have no idea what it even means and I can’t delete it
I feel like it has been slowly getting worse for years, but it has gotten way worse in the last couple months. This could just be my extreme anti-AI bias but it feels like since they introduced “apple intelligence” autocorrect cratered. I also use voice to text a lot on my phone, and I have to manually correct probably 1/4 of what it spits out. That’s always been bad, I’ve always had to correct it, but it has also gotten way worse in the last couple months.
Decided yesterday I can no longer use that function. It has become way too unreliable! The words it was putting in was offensive! Words I’ve never used! And if I find something offensive, it’s bad!!!
Ok, so this is a thing. Any ideas what’s happening? Just a strange bug that will be fixed soon? Or is this the beginning of enshitification of many systems on our phones, laptops etc. due to AI, people being fired or something else?
One thing I could imagine is if there’s some sort of feedback loop between LLMs processing information (and „talking with themselves“ in their own AI language) that is somehow connected to autocorrect. If this was the case, I can easily see the sheer mass of weird new language information overwriting actual human languages and what we have fed into the system so far.
What’s strange is how it seems to happen on both IOS and Android. I’m mostly using Android, but I’ve also noticed it on my wifes Iphone.
But maybe they both rolled out a similar feature recently.
Perfectly possible, if LLMs continue to be trained on all available content, which will of course include LLM-generated content (unless there’s a very clever way of filtering LLM content out … a good job for an LLM of course … oh, wait … )
Lol, reading back on my post, it begs the question: are we Elektronauts finally ready to cook our own conspiracy theory, like every online community is supposed to at some point?
I’ve noticed it on iPhone too, it’s so much more aggressive in “correcting” ordinary words, and yet not any better at telling well from we’ll in context, blaming sloptech hype