Introducing Hyperspace: A Decentralized AI Network for Diverse Model Experiences

Meet Nicholas Schlapfer, an AI engineer introducing Hyperspace, a decentralized AI network that empowers users to contribute their personal computer resources and unlock innovative applications in AI engineering.

  • 1. Speaker: Nicholas Schlapfer, an AI engineer at Hyperspace.
  • 2. Hyperspace is a decentralized AI network with no GPUs.
  • 3. The company builds and relies on a community contributing resources from their personal computers to the network.
  • 4. Hyperspace has a product called AIOS, available for Windows and Mac, which uses LlamaCpp for inference.
  • 5. Hyperspace aims for diverse models instead of relying on one large closed-source model for better AI experience.
  • 6. The company is working on a new self-titled product, Hyperspace, built on their decentralized network.
  • 7. Hyperspace is a mix of prompt engineering, visual React Flow, Python execution, and "rag" (possibly web browsing).
  • 8. The first goal for Hyperspace is to build a fine-tuned model that outputs a genetic planning experience.
  • 9. Users can input a query and receive a directed acyclic graph (DAG) or JSON format, representing a methodical plan from the query.
  • 10. Hyperspace includes an in-house web scraping experience for learning management systems (LMS), using Puppeteer and Beautiful Soup.
  • 11. The web scraping converts HTML to markdown for easier LMS digestion.
  • 12. Hyperspace has a node editor in a terminal for power users.
  • 13. The node editor allows changing each node in the React Flow according to user needs, based on the DAG orchestration model.
  • 14. Users have full control over nodes, including editing titles, descriptions, expected output, and adding nodes as needed.
  • 15. Hyperspace utilizes a reasoning model and summarization model for synthesized answers, using Quen 2 Instruct and Llama 370b.
  • 16. The system provides diverse sets of outputs with integrated Python execution in the terminal.
  • 17. Hyperspace aims to lay the groundwork for "agentic behavior" by offering core primitives needed for an agent, such as a virtual file system, memory management, and code generation.
  • 18. The new Hyperspace product will be available later in the week via Weblist.
  • 19. Keep an eye on Hyperspace AI's Twitter account for updates on the release.
  • 20. Hyperspace believes in diverse models to provide the best AI experience, combining various experts instead of relying on a single large model.
  • 21. The company values community contributions and resources provided by individuals through their personal computers.
  • 22. Hyperspace's AIOS product offers inference capabilities from various sources, enabling conversational experiences like chatting with someone in Belgium.
  • 23. Hyperspace plans to continuously improve and expand its offerings within the decentralized AI network ecosystem.
  • 24. The new Hyperspace product combines multiple technologies and techniques to provide a unique and innovative AI experience for users.

Source: AI Engineer via YouTube

❓ What do you think? What are your thoughts on the ideas shared in this video? Feel free to share your thoughts in the comments!