Chaotic Generative AI Group Chat: Art Generation Workflow, AI Tools, and Learning

Topics Discussed in Chaotic Generative AI Group Chat Transcript:

AI Art Generation Workflow

  • Discussion on using Stable Diffusion for AI art generation.
  • Workflow involved using text2img with controlnet depth map of the image.
  • Depth map was obtained from Clipdrop’s API.
  • Automatic1111 was used for the entire process.
  • Iteration loops were run in every step until satisfied with the output before moving to the next step.
  • Retakes are common in AI art generation.
  • Fascination with how a deterministic input can give infinite non-deterministic output.
  • Non-deterministic output still stems from a classical random number generator.
  • The concept of seeds is simple but quite useful.
  • There are infinite seeds in theory.

AI Tools and Plugins

  • Discussion on Langchain agents and ChatGPT plugins.
  • Certain things are easier and more fun with a UI.
  • Depth maps provide more control.

AI Learning

  • Suggestion to try few-shot learning for training.
  • Ran out of context length before stuffing enough context for significant improvements.

Humor

  • Ran out of context length before stuffing enough context for significant improvements 😂
  • This debate itself is infinite 😆
  • Sudharshan’s tweet: https://twitter.com/sudharshan_/status/1441267649479470080

The description and link can be mismatched because of extraction errors.

← OpenAI's Development Trajectory and the AI Dilemma
AI Model Customization and OpenAI Plugins →