site banner

Friday Fun Thread for July 21, 2023

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

3
Jump in the discussion.

No email address required.

I was trying out Bing Chat's GPT-4 derived multimodal feature to transcribe a handwritten table in a note taking app.

It worked the first try, and did a pretty good job, but when I tried to reproduce its success, with the same prompt and image later, it absolutely shit the bed and began hallucinating based off random UI elements in the screenshot.

Huh. I mean, I've show that it can do it, even with my doctor's handwriting (it's not that bad OK?), now I need to figure out how to do it consistently.

I initially had to redo it because copy pasting broke the markdown formatting Bing was using, and I still haven't figured out a way around it.

Give the google drive/docs scanner a shot also; I've used it for the last couple years with great success. I'm told it's hit and miss for other people though.

I didn't want just OCR, especially since I wanted to preserve tables I'd drawn by hand, but sure I'll check it out!

It preserves my indents and margins if I scan the whole paper, but I've never tried it on a hand drawn table.