HOME

Date: [2023-03-19 Sun]

Compression is AI

Table of Contents

1. Thoughts

  • Not all observations remain as statements (Inferential Learning) in our brain.
    • We see, then capture according to what our brain understands as important at that time and we discard rest of the information
    • E.g. Cartpole problem: (All claims are "I think so")
      • We don't remember the whole image
      • We don't remember the whole sequence of positions and angle
      • We remember some state (position and angle) and the nature of their transition (e.g. monotonic decreasing)
        • We have seen such transitions many times before. So it is easy/natural to compress the input in those terms.

You can send your feedback, queries here