Zero-Shot Image Captioning Allowing Dialog Language Models to approximate the content of an image through Image Captioning with the help of CLIP. Overview We allowed Language Models like GPT-J 6B to be able to have an approximate overview of images by classifying an image with labels using natural language. Our implementation was heavily inspired through Google Research’s Socratic Models which aims to provide multimodal reasoning for Large Language Models such as GPT-3.

Announcing DNN (Definitely Not Nitro) TL;DR: The developers for Not Quite Nitro have a very misleading privacy policy that admits that their bot, Not Quite Nitro, stores sensitive information about users in the case of internal errors on their end without the user’s knowledge. We did not like their policy or practices, so we decided to make an open-source alternative. What’s the Problem? We have a Discord server where user privacy is sacred.