Manifold is a distributed research group working on significantly improving AI systems and applying them to the toughest problems humanity faces. We do this by advancing theory and also building exceptional open source tools. We work in a different way to most other organizations, combining the strengths of focused research labs and large open source organizations. We also take a capabilities focused view of AI, and believe intelligent systems should have:

• Multimodality: Learning systems should be able to handle multiple objectives across multiple modalities.
• Continual Learning: They should be able to update and adapt to changing data over time.
• Modularity and Composability: Composability & Modularity leads to ease of reuse, fast adaptation, and human interpretability.

We think if the above is achieved, there are numerous near term gains in various high impact applications.

• NEKO 🐈, an open source effort to build a massive generalist model capable of performing dozens of tasks across image, text, and control modalities.
• AgentForge ⚒️ aims to build generic tools and models to rapidly build Intelligent Agents capable of digital tool use.

The best way to get plugged in and contribute to the research and development happening at Manifold is to:

1. Check out the above project links, which include roadmaps, contribution guides, issues and project status, and more.
2. Join the conversation over on the Manifold Discord, where we do all our coordinated work.
3. If you want to be actively involved (~10 hrs/ week), consider joining the OS Research Team! Check out what it means to be a OS Research Team member and the associated expectations here.