Theseus - The Mouse in the Maze

I built a mechanical mouse called Theseus that could navigate a maze using telephone relay circuits and magnetism. This was one of the first machines to demonstrate learning behavior.

How It Worked

The system used:

1. Telephone relay circuits: To store and process maze information
2. Magnetism: Steel wires embedded in the maze walls guided the mouse
3. Learning circuits: Adaptive behavior through relay logic

The Learning Process

1. Initial exploration: The mouse moves randomly through the maze
2. Success detection: When it reaches the goal, the path is “remembered”
3. Memory storage: Relay circuits store the successful path
4. Future navigation: Subsequent attempts use the stored information

The mouse would first explore randomly, but after finding the goal once, it could navigate directly to the goal every time.

Technical Details

• Relay memory: 40+ telephone relays stored the maze solution
• Maze dimensions: 5x5 grid (25 possible positions)
• Navigation: Magnetic field sensors detected walls
• Size: About the size of a small cat

Significance

This was one of the first concrete examples of machine learning - a machine that could:

• Sense its environment
• Remember successful paths
• Improve its performance over time

The project demonstrated that machines could exhibit behavior that appeared intelligent, sparking discussions about the nature of intelligence itself.

Reception

The machine was demonstrated at Bell Labs and generated significant interest. It was featured in Popular Science and Scientific American, making it one of the first AI projects to receive public attention.

---

Theseus showed that intelligence isn’t unique to biological systems - it can be reproduced in machines using the right circuits.