I’ve seen this with gpt4. If I ask it to proofread text with errors it consistently does a great job, but if I prompt it to proofread a text without errors, it hallucinates them. It’s funny to see Microsoft having the same issue.

I’ve found SwiftKey predictions have gotten super terrible recently. They used to be very good. I blame AI.

Apple has gotten terrible. If i don’t proofread it will change words to what it thinks they should be without prompting me. Also, editing sentences on iOS sucks. They changed that too. That way after they fuck up my sentence it’s also way harder to go back and fix it.