Multi-species skill-evolution: stages and assumptions
The fundamental assumptions for the whole project:
- Skills are self-assembled, i.e. without divine or other programmer
- All skills rely on neural, dendrite, and synapse programming. They
therefore use extremely slow processors Neuron run at 100Hz = 10ms, more than 10 million times slower than a PC, but use billions of processors in parallel compared to 4 or 8 for a PC.
- Many skills are not inherited, i.e. programed by the genome. An example
includes the ability to speak English and French. Such skills have been
learnt after conception and birth.
- Almost all skills have evolved, including learnt skills.
- Innate and learnt skills have separate mechanisms for evolution.
- Cooperation plays a role in the various mechanisms for evolution.
- The functionality of skills accrues over evolution and maturation, i.e.
later skills draw on earlier skills.
The framing assumptions for the whole project, separating skills from other components:
- Input from the world to skills comes through the perception system, primarily vision and hearing.
- All data from the perceptual analysis is available simultaneously as input to skills, somewhat like a buffet or smorgasbord.
- In some cases, skills and muscle action may be required for perception, such as for moving and focussing the eyes.
- Output to the world goes through action, through the motor-cortex,
muscles controlling the angles of joints connecting limbs.
- Feedback controls may link perception to action without going through skills.
- A skeleton with 15 joints and 3 polar coordinates per joint is used for modeling locomotion.
- Joint angles are used instead of relative muscle-tension, to simplify the model
- Neurons are simplified.
- 25 frames per second equals 40 msec between frames, which is used as default neuron firing rate with a synchronized clock
- Numeric instead of binary values.