The MUHAI project introduces meaning and understanding in artificial intelligence
Responsible human-centric AI needs a way to deal with meaning. The MUHAI project tackles this foundational question by developing computational models of narrative construction. Creating or understanding narratives requires the integration of information coming from sensory-motor embodiment, measurement data, language, semantic memory, mental simulation, and episodic memory.
MUHAI uses two domains as testbeds for the development of the required breakthroughs: common sense pragmatic knowledge about the world needed for the domain of cooking and social knowledge needed for coming up, understanding and checking data stories about inequality in society.
The project relies on many existing techniques of AI ranging from deep learning networks to knowledge graphs, but will push their boundaries and develop new techniques all operating in the service of giving AI systems a better grip on meaning and thus on explanation and other key issues for achieving human-centric AI.
The outcome of MUHAI is twofold. It will push the state of the art in cognitive home robotics, particularly for food production and the management of food resources, and it will provide tools for social scientists to better understand social phenomena, as for example the persistence of inequality in our society.
The MUHAI project has started in October 2020 and will finish 48 months later.