Million Minds

Phase 1: Implement basics

Start simple with a text-only system

• Sensor (Text Retina)
  Convert raw text into multi-scale feature SDRs using biologically inspired “retina” patches.
• Thalamus
  Gate sensor SDRs and mark landmarks based on instability and morph flags.
• Cortical Column (L4 & L2/3 Prototype)
  Implement a single-column cortical model that learns temporal transitions and pools stable features.
• Column Stack Demo
  Run an end-to-end simulation of text reading using the cortical column and thalamus.

Phase 2: Add remaining components

• Pose System
  Add sensorimotor grounding via 1-D grid-cell–like modules.
• L2/3 Pooling Upgrade
  Replace simple pooling with leaky, gated temporal pooling.
• Lateral Bus
  Voting between neighboring columns for consensus alignment.
• Persistence Layer
  Save and reload learned memory and pooled chunks.
• Variable saccades
  Allow jump size to change based on error/surprise.

Phase 3: More advanced features

• Meta Layer
  Higher-level columns to chunk smaller features into larger concepts
• Memory
  1. Episodic
  2. Semantic
  3. Procedural
• Query/Dialogue
  Interact with the system, e.g. ask questions

Phase 4: Other modalities

• Images
  1. Text via images
  2. Images
• Videos
• 3D Virtual Environment
• Hardware sensors