#art #ai #llm #transhumanism Created at 2022-12-12 # [Anonymous feedback](https://www.admonymous.co/louis030195) # [[Epistemic status]] #shower-thought #to-digest # Related - [[Blenderbot3 GPT3 DALLE stable diffusion raspberrypi conversation AI]] - [[How to - Digging specific semantic space together]] - [[Seeker search augmented conversational bot]] - [[Conversations augmented by text-to-image AI]] - [[GT3 - lexicat art - stable diffusion]] - [[Obsidian plugin or templater ideas]] # TODO > [!TODO] TODO # Multimodal AI self-chat - BlenderBot3 GPT3 stable diffusion lexica art ![[Augmenting your brain with AI multimodality - Example]] ![[Augmenting your brain with AI multimodality - Example2]] Purely random/funny/useless tech thing:I've been playing with [https://lexica.art](https://lexica.art/) API lately with obsidian.md, basically insert stable diffusion images around my thoughts (it's fast unlike actually generating with stable diffusion)In the past I've played a lot with [https://parl.ai](https://parl.ai/) chat bots, and you can do self chat (i.e. two bots talking to each other). There are bots that search the internet in real-time to enrich their conversations ([https://louis030195.medium.com/deploy-seeker-search-augmented-conversational-ai-on-kubernetes-in-5-minutes-81a61aa4e749](https://louis030195.medium.com/deploy-seeker-search-augmented-conversational-ai-on-kubernetes-in-5-minutes-81a61aa4e749)) (maybe chatgpt does it also, but BlenderBot2,3 and Seeker does it for a while) My idea is to have two bots talking to each other and in parallel query lexica.art given the conversations (and maybe ask a language model to turn the conversation into a lexica art query, but that get too complex RN) and stream the multimodal conversation on twitch or whatever streaming website[https://parl.ai/docs/tutorial_tipsntricks.html#self-chats](https://parl.ai/docs/tutorial_tipsntricks.html#self-chats)Thoughts?