Inspiration from nature to design computing systems

^^^Lesson #1: Emergence: The whole can be more than the sum of its parts.

You have to starve the ants to get them to do anything. Scent trails lead swarming. Simple but not robust optimization: Ants collectively select the shortest path to the food source. First ant that returns to the next more likely to find shortest path, then reinforced through feedback.

Highway effect: harder to change paths, but robustness through evaporation corrects

Examples:
-Travelling Salesman Problem (TSP): what is the most efficient path to visit prospects. Traveling sales-ants (lots of math here):
*Agents build possible tours
*Agents reinforce
*Virtual pheromone evaporates
-Routing in communications networks (95% in throughput and packet delays)
*Simple agents launched to go from a source to a destination node
*Agent updates routing tables on its way to its desitnation, viewing its source as a designation
*Agent infuence decreases with age
*Agents are artificially delayed at congested nodes

^^^Lesson #2: Simple rules rule

-Bucket brigades in harvester ants, organized by ant size
*first and smallest collects seed and starts to carry until it meets larger worker. Larger workers more efficient, their use is optimized.
-Bucket brigades at Taco Bell: (34% increase in productivity)
*organizing labor by increasing efficiency
*least efficient starts task, most efficient finishes task
*simple rule for division of labor increases efficiency
-Protector-Aggressor Game (greek fonts prevented this demo)
-Bad News
*Hard to predict aggregate outcome
*Small changes have large outcome effects
*People are a part of something bigger, but they dont know what by looking at the rules
-Good News
*Outcomes can be modelled bottom up
*If you can predict what will happen you can design the rules to effect behavior
-Southwest Airlines case
*Problem: Optimized cargo routing using smple rules
*Results: 71% improvement, $10m/yr
-A cautionary tale about blindly following simply rules
*Ants can follow each other in a circle until they die
*More pheremone there is the more the lay and the faster they go, and then die of exhaustion

^^^Lession #3: No One Needs to be in Control

-Cooperative transport: push and pull in all direction until by chance direction is established, robots designed with similar principles
-Nest construction in wasps (building based upon what is seen locally)
*agents move randomly on a 3d grid of sites
*an agent deposits a brick every time it finds a stimulating configuration
*feedback

^^^Lession #4: Size Matters

-Water-handling task partitioning
*Polyvalence in smaller colonies (when smaller everyone knows everything)
*Specialization in larger colonies (can't know everything)
-S-curve with system efficiency on vertical axis, system size on horizontal
*Swarming at point of increasing and decreasing returns

^^^Lesson #5: Lessons of Prey

-How do we shape emergence?
-How do we have emergent behavior by design?