I590 - Artificial Life
Presentation - Analog
Attach:presentation.ppt Δ
Mills presented previously on function approximation; Yaeger would like to see more neural network-oriented applications.
Ideas
TODO: Need to update this.
- Information - static systems tell you nothing because the state was already known. Also, randomness tells you nothing either because there is no coorelation.
- Complexity lives between the chaotic and the predictable.
- Evolution and the 2nd law of thermodynamics are observable processes, not driving forces. They are consequences of more fundamental phenomenon, whatever those are.
- There are CA configurations you cannot predict (the halting problem).
- Do physical constraints on neuronal efficiency hint at the existence of an upper bound on intelligence?
- Learning - We "learn" a lot of junk, but it quickly dissipates because it is essentially random; only environmental invariants (or other interesting phenomenon) are reinforced -- memor church.
- Do ALife researchers believe their systems to be value-neutral?
- It is hard to determine the genotype by observering the phenotype.
- However, GAs generally encode specific phenotypic traits. Does this not presuppose the meaning of genese?
- Where did the "meaning" come from in biological systems? If evolution is a blind agent, it seems wrong (on a philosophical level) to predetermine phenotypic traits, even within a range, from the genotypic encoding.
- The arbitrary selection of traits to represent and traits to omit is inherently biased -- you get what you're looking for.
- Why litmus tests for intelligence (The Planet of the Apes)?
- Emergence - the exposure of indirect variables; the tendency of a behavior to emerge (e.g. intelligence), aka "resilent emergent properties" -- Does this say that emergence is the behavior of variables that cannot be directly manipulated (phenotypic complexity)? Could this be physical basis for tacitity? Where does this tacitity come from?
Points to follow up on:
- The ALife tautology in relation to Darwinian assertions.
- Business modelling and ALife.
- Debating ALife with Mike - What does it mean to be random?