Ambiguity is when the same input or situation could mean several different things. The world is full of ambiguity including natural language, images, and sensor data. Sometimes ambiguity can be resolved based on internal coherence of the data, fusion of multiple modes of sensing or the same sensors over a longer time. Often there is some residual ambiguity typically expressed as a set of alternatives with a confidence measure for each, but sometimes a 'best guess' approach is needed. In bottom up systems, ambiguity at a low level, for example individual words in language may need to be resolved at a higher level, such as the meaning of a sentance. Top down systems may be able to push down expectations to lower levels, for example a fixed vocabulary in a home automation system, that makes low level disambiguation easier.
Scene with ambiguous labelling.