The MUHAI project introduces meaning and understanding in AI
The new Pathfinder project MUHAI is funded by the EU Commission (EIC-FETPROACT-05-2019) to develop foundational technologies for achieving human-centric artificial intelligence. Human-centric AI aspires to empower rather than replace humans, which implies that it should be able to explain and defend how it reached decisions, can take advice from humans, and is compatible with the ethical and moral standards that we expect from agents in our society. Human-centric AI contrasts with the data-centric AI approach that is now engulfing industry and society. Data-centric AI uses big data and statistical machine learning to make decisions by exploiting statistical patterns. These patterns are often coded in ways which are inaccessible to humans, including the developers of these systems. They make the promise to overpower humans and to make us superfluous, raising many ethical and legal questions. The project is based on the idea that in order to be human-centric, AI must learn how to cope with meaning and understanding. MUHAI new approach will be tested in cases that require the use of common sense about everyday activities such as cooking and an understanding of social phenomena. MUHAI objective is to push the state of the art in cognitive home robotics while providing tools for social scientists to better understand society.
Why are the neighbourhoods in some cities sharply divided along income boundaries, while in other cities not? Was this always the case in different periods of history? And in different cultures? Has social mobility increased or decreased over time? Why does life expectancy correlate with income?
If the proof of the pudding is in the eating then the ultimate test for understanding an instruction is its proper execution. This view greatly expands the scope of natural language understanding beyond the usual syntactic and semantic analysis. In this part of the MUHAI project we seek to operationalize the basic principles of human-centric AI so that machines will be able to understand how to perform everyday actions in the cooking domain.