We are still in the early days of AI assistants and many things still need to be invented. Our research is driven by real challenges and datasets, and its primary aim is to let makers expand the capabilities of their AI assistants.
AI assistants should be able to help with more than one task. How can we re-use experience in learning one task to learn another faster?
Training data is a bottleneck in every conversational AI project. How can we build models that can learn from just a few examples?
How can we build sentence embeddings that are sensitive to the rules of conversation?