I'm still caught up in Area51 trying to get a proposal together for AI and robotics. Currently this proposal is floundering as there are several small committal phase proposals (Machine Learning, Computer Vision, Robotics, NLP/CL) that would be covered by my proposal and I would imagine many more that are. The problem is that too many subjects overlap causing squabbles over which discipline should own a topic, so they branch out in small numbers. The less fortunate topics are left to live in false hope until eventually the inevitable beta doom hits them and they slide down the snake back to the start.
The original AI proposal was canned because of the honeypot-effect and it's multi-discipline nature. Machine learning would be covered by stats, Robotics by the electrical crowd etc.
It has occurred to me that the [SE] model cannot cope with handling cross-topics and that the real expertise will be lost in corners of unheard of sites or lost altogether.
A perfect example of overlap is illustrated in Why create a Mechatronics Proposal
Is there a fundamental problem with overlapping topics in [SE] that requires everybody to step back and re-think how this whole thing works, or is there a way we can unify topics into an [SO] magnitude site but allow cross-polination of q and a's?