Stephen Brooks 🦆 on Nostr: I've seen LLMs use image-to-text and speech-to-text as input, so maybe that's where ...

I've seen LLMs use image-to-text and speech-to-text as input, so maybe that's where these funny mistakes are coming from?