The Shitposting AI - An Ironic Solution to Fruitless Online Debate
summary
Claims an automated system can replace humans in comment threads by training GPT2 variants on scraped Reddit and Twitter data and running continuous fine‑tuning. The prototype uses tokenization, a server-hosted generative model (later upgraded to F2), TTS (initial Tacotron‑style attempts then Cookie TTS or cloud services like Amazon Polly/11 Labs), Raspberry Pi + Arduino glass robot heads for speaking; it mostly works for demos but hallucinates links, produces NSFW noise, and is legally risky so it is not released.
AI-written summary · about
chapters (16)
- 0:00Shitposting AI project intro
- 4:03Problem with online comments
- 6:10AI bots automate replies
- 7:36Rehumanizing AI with robot heads
- 8:18NLP for comment generation
- 14:43Social media comment generator
- 21:33Choosing AI model
- 23:44Training data from Reddit/Twitter
- 26:58Training neural network
- 28:25AI voices for comments
- 31:20Robot heads hardware
- 33:07Assembling Shitposting AI
- 36:40Live AI demo
- 37:01Why project was unreleased
- 38:05Other AI innovations
- 38:46Q&A
related talks
01h 03mIt's All About AI
09mRethinking AI: Building What Matters, Not Just What's Possible
41mThe Value of LLMs - Why Stupid Machines Are Smarter Than You Think
38mGenerative AI Agents: How they Work and How to Build One
51mOpenAI and the AI that may overtake humans
45mWeb Development in the Age of AI
46mConversations with ChatGPT: illusion or reality?
25mFrom Treatise to Tech: The Development of ChatSOC, an AI Research Assistant