Non-Adaptive Evolvability

on YouTube ↗46m

" summary

Uses Avida digital organisms to show natural selection drives mutation rates down on rugged fitness landscapes, producing order-of-magnitude lower mutation rates than the long-term-optimal fixed-rate benchmark; this holds across recombination, self‑referential rates, population sizes and changing environments and fails only on smooth landscapes or with implausibly large populations. Separately, evolves networks on an 8‑pixel retina task and finds a connection-cost penalty yields highly modular architectures that transfer and learn new tasks faster and avoid catastrophic forgetting, whereas selection for performance alone produces tangled, less-evolvable solutions.

AI-written summary · about

" chapters (8)

1. 0:00Introduction
2. 0:21NonAdaptive Evolvability
3. 1:47Mutation Rates
4. 2:02Experimental Approach
5. 18:19Modularity
6. 22:17Evolutionary Performance
7. 30:23Biological Implications
8. 33:20Evolution