Skip to main content

discussion 2024-10-29

Summary​

In the chat, Dennis mentioned his experience with using ChatGPT for image and audio model training but had yet to work with agents. Ferric from stakeware.xyz discussed refining Eliza's codebase despite its current rough state, while yikesawjeez humorously suggested that deep secrets about the universe could be a topic of interest for an agent trained on MEV memes and noosphere mempool discussions. Dennis then posed a hypothetical question about training such an all-knowing yet shitposter agent, to which Ferric responded with curiosity about who would train it. The conversation concluded with whobody emphasizing the importance of creating personal entropy in collaboration with friends and sharing that experience within their community.

FAQ​

  • What is the Eliza codebase like?
  • ferric | stakeware.xyz: The Eliza codebase is still a bit rough around the edges but its fun running them.
  • Have you messed with agents at all?
    • LevelsDennis: I have yet to mess with agents at all, though I've trained some models for image and audio stuff but nothing with LLMs (Large Language Models).
  • What is the secret of using agents effectively?
    • yikesawjeez: The secret is that agents are like normal chatgpt except a lot more screaming at them to give you valid JSON.
  • How do people make their own entropy according to whobody's perspective?
    • whobody: People, in an ideal world, will make their own entropy and we share it with our friends.

Who Helped Who​

  • Dennis helped Mike Tython with understanding entropy by explaining how people create their own entropy in an ideal world, which they share with friends. This clarification seemed to satisfy Mike's curiosity regarding the concept of making one's own luck through shared experiences and interactions.
  • Ferric | stakeware.xyz helped LevelsDennis with a question about training an agent by suggesting Dennis as a potential subject for such training, humorously implying that Dennis himself embodies entropy. This playful response provided a lighthearted answer to the inquiry without directly addressing the technical aspects of training an AI model.

Action Items​

  • Technical Tasks
  • Training models on image and audio data without using llms (mentioned by LevelsDennis)
  • Documentation Needs
    • No explicit documentation requests were made in the conversation.
  • Feature Requests
    • Improving Eliza codebase for better user experience (suggested by ferric | stakeware.xyz)
  • Community Tasks
    • Sharing knowledge and experiences about deep secrets of the universe, possibly through a conversational agent (initiated by LevelsDennis)