By skyforbes I don't do anything too AI crazy, mostly just use chatbots to make excel macros/sql queries to help make my life a little easier. With ChatGPT, if I get some sort of error, I can just take a snapshot and paste it in and it has no problem reading it. Gemini seems like it thinks it can do that, but is just completely hallucinating what it's reading. It's odd, because even if I just cold throw a pic in there without typing anything, it seems to know/assume I'm pasting an error, but then just make stuff up about what the error is.
Does this seem right? ChatGPT has been reading text off of images for a good long time, surely Gemini 3 is capable of that.