Previous Page Table of Contents Next Page

The third approach focuses on evolution, on simple brain-based roots for skills and on the sequence leading to the target skill

This approach starts with the Darwinian theme that skills evolve. If we trace how skills evolve it might help us understand how skills function. Even very complex skills must have evolved from very simple roots. Ideally we have chains of skills evolving toward the skill we want to work with. The root skill and the skills in the chain presumably are simpler and therefore easier to simulate, especially if we track evolution backward over many species. We assume that there is no discontinuity in evolution, i.e. that all skills emerged in a series of small steps.

To use this approach we have to show how each step in the evolution of the skill adds to Darwinian fitness, i.e. increases the likelihood to have offspring and to somehow carry forward the given variant of the skill to the next generation. This goal of finding and moving forward from the simple roots of complex skills lies at the core of this research project. Ideally we can project the evolution of present-day skills into the future.

The second starting point is to assume that skills represent information-processing algorithms that are stored in the brain, and that are processed in the brain to produce skilled behaviour. We therefore have to show how each successive version of the skill can be stored and processed as brain-compatible information-processing. The brain-processing assumption adds constraints on the speed of processing that make it difficult to model skills with super-computers to achieve real-time performance. This starting point provides a second motivation to go back to the roots, to find simple enough skills that we can model successfully.

The third starting point is to to look for species that don't present great difficulties in projecting the skill forward. We need a skeleton, joints, and muscles system that is sufficiently similar to that of humans. We also need a perception system with eyes and ears that are fairly similar.

The approach, then, consists of a whole series of simulation models of skills that could plausibly be executed by the brain and that could fit a reasonable chain of evolution.

How many chains of skill-evolution do we need to model?

We would like skilled actions that have analogues in the sequence to human skills. For survival and for reproduction we need locomotion. Starting with vertebrates, and then mammals, the skeletons with limbs, joints, and neuron-activated muscles are similar. For most of these species sensory perception is similar wih eyes, ears, and spatial orientation. We can therefore assume similar input and output and focus on the information-processing algorithms (skills) of chasing and escaping.

Cooperation, communication, and learning are important for skill-evolution, so we will look for relevant skills such as mimicry.