A magical emergent algorithm: Swarm Sifting Sort.
- A magical emergent algorithm: Swarm Sifting Sort.
- This algorithm works even with extremely noisy input.
- The magic is it requires no coordination or top down control.
- All you need is:
- 1) A consistent bias for each action that moves each item closer to its correct position.
- 2) An authentic signal that has no structural incentive for cheating in each action.
- 3) Lots and lots of actions: the more, the better.
- As long as you have these, it doesn't matter how noisy the signal is, over time the emergent algorithm will converge to the correct result.
- The larger and more active the swarm, the faster the sorting.
- The noisier the signal, the larger the swarm you need.
- If you have a massive swarm it doesn't matter how noisy the signal is.
- A lot of search ranking techniques reduce to this technique.
- Here's another example for moving items in a warehouse:
- When an agent is walking by an object, if the object wants to go in the direction the agent is walking, pick it up.
- As soon as the agent's incremental step will move the object farther from where the object wants to go, set it down.
- That's it!
- This is easy for robots like Kiva robots to do leading to emergently sorted warehouses, but it's also plausible for humans if they could quickly determine where an object they were passing by needed to go.