I had posted a discussion to Reddit[0] about a year ago that touched on a similar idea. Its a model for AI using a community of simple IA that as a whole becomes the AI. It injects social behaviours into the IA to govern how they interact with the community and to allow bias to develop among groups, and domain proficiencies to develop within individual IA. Because I envision the model being distributed, the proposed implementation includes a blockchain-like ledger of points of confirmed knowledge. I'm really not all that knowledgeable on the subject of AI but given the relevance and knowledge of HN I figured I'd toss it here to see if anyone wants to read a wall of text and give feedback. I have a backlog of books on AI I want to read, so if there's anything you'd like to add to my reading list after skimming through the idea I'd appreciate any suggestions :)
[0] http://www.reddit.com/r/artificial/comments/1t5525/community...