Sandbox environment to create self-replicating machines thru natural selection. This is ongoing project without a release yet.
See wiki for details.
Abiogenesis theories postulate that there was a state in early Earth which allowed self-replicating molecule to appear. One of the fascinating aspects of such idea is that because of the errors in replication slightly different copies of such molecule were created and from that point on natural selection begins. One of the results of such natural selection is human intelligence.
One of the interesting points of view on human intelligence (need links) is that human understanding of the world is subjective - we're biased on our ways of perception, our biological background. What that means is we can't expect rational behavior if place human intelligence into conditions which don't have anything to do with human experience.
If we connect those two ideas one can wonder if one of the ways to achieve strong artificial intelligence could be to create specific environment which allows self-replication and let natural selection do the work to create intelligence which makes sense in that environment. There are numerous yet unanswered questions if we'll go in that direction of course and this project makes a step in order to answer some of those.
- What are minimal requirements for environment in order to create self-replicating mechanisms?
- Are self-replication and randomness the only key elements for starting natural selection?
- How to detect self-replication once it occurs in created environment?
- Could natural selection in very limited artificial environment lead to creation of intelligence?
- Is there a way to optimize that process to get to intelligence faster comparing to biological evolution?
- What are the key requirements for the environment with self-replication to achieve intelligent actions?
- How to define if one's reaction is intelligent in artificial environment?
Whateley project's main aim is to define the minimum rules needed to create an environment in which self-replication is possible and once achieved is dominating. Thus it focuses on the first three questions.