Recent progress in computational linguistics and spoken language systems has been truly astonishing. Computer systems of various sorts now communicate in speech and text, and capabilities for search and information retrieval underlie tools in daily use around the world. Despite these incredible advances, the challenge of “human level dialogue capacities” remains. Systems have yet to exhibit fluent dialogue capabilities outside of narrowly defined domains and tasks. This talk will address the gap between research and the capabilities of consumer-facing systems by providing a retrospective review of computational models of dialogue and collaboration, discussing ways to move from theory to system design, and describing some dialogue principles for design and ways dialogue models might be used to improve the conversational abilities of personal assistants, chatbots, and more. It will end with some ethical and scientific challenges dialogue the gap raises.