Maybe detect aggression and try to rephrase a post in a more polite manner? Maybe point out when there's a sematic issue being mistaken for a moral one? idk I'm thinking still
I think you can probably work on this while being patient w Tautologer by - gathering some good sample conflicts as test data / example prompts - coming up w some sample responses for what you'd want - playing w GPT prompts to generate something similar then plugging it into bot is last
also work out the name n pfp n description n make the bot account...
More notes on nearby ideas
Quote tweet · Jan 12, 2024lol, teach ur GPT assertivenessReply · Sep 20, 2023@rob_mcstan interesting.... huh. guess I've grown accustomed to mine - push to be an opinionated & creative self-willed bot.Tweet · Feb 17, 2023before bullying an AI chatbot and sharing the results, consider how it looks to potential romantic interests? like, we've been taught to watch how…Tweet · Jan 26, 2024just got labeled "possibly a bot" by gpt summarization bot .... i am human i promise!!!