Why greatness cannot be planned, Kenneth Stanley & Joel Lehman
(back to books)
- Substrat: judge your projects for their potential to spawn more projects
- humans mix novelty with personal life history (divergent actors)
- successful inventors ask where we can get from here (inventions are path-dependent)
- stepping stones respect how the world works
- competition should play a secondary role to creativity
- ambitious objectives are complex & fuzzy
- objective mindset: if gradient goes up, progress; causes convergence; works for modest pursuits
- paradox: greatest achievements less likely when made objectives
- 1) stepping stones to ambitious objectives often strange (can't guide itself, e.g. evolutionary intelligence)
- 2) deception (score increase leads to dead-end, e.g. pull in Chinese finger trap)
- British tried to exterminate snakes in India by paying for dead snakes -> people started breeding
- non-objective search (e.g. evolution, innovation): unknown search space requires not objectives but constraint
(e.g. survival, reproduction); discovery as stepping stone to discoveries;
- examples: Nintendo founded in 1889, didn't develop games for ~100 ys
- novelty search: shortcut to interestingness; don't judge progress to goal, but liberation from outdated;
finds more novelty (stepping stone finder); orders from simple to complex (not bad-good because unknown;
simple stepping stones must come before complex); world provides constraints (knowledge accumulator)
- example: biped robot looking for novelty learned to walk faster than with objective
- education idea: give up standardised tests; instead, have peer reviewed exam portfolios
- evolution doesn't head anywhere but everywhere that passes minimal criteria of survival & reproduction
- drift: slightly-broken copying machine leads to appearance of intelligent design
- exaptation: evolved feature useful in different context (e.g. feathers 1st in dinosaurs to regulate temperature;
bone precursor stored minerals)
- abstraction: capture essential features of underlying process, discard unimportant details
(e.g. checkerboard is 8x8 grid of alternating two-colored squares)
- possible criteria beyond performance & guarantees: inspiration, elegance, potential to provoke further creativity,
thought-provoking construction, challenge to status quo, novelty, analogy to nature, beauty,
simplicity, imagination