Intelligence without representation
Artificial Intelligence
Cognitive situated agents learn to name actions
Proceedings of the 2006 conference on ECAI 2006: 17th European Conference on Artificial Intelligence August 29 -- September 1, 2006, Riva del Garda, Italy
Model Suites For Multi-Layered Database Modelling
Proceedings of the 2010 conference on Information Modelling and Knowledge Bases XXI
Proceedings of the 2010 conference on Information Modelling and Knowledge Bases XXI
Hi-index | 0.00 |
Here is considered emergence of language in an environment, where agents communicate about issues which are not observable at the moment of communication, e.g. they search food and exits in a labyrinth and communication helps them to achieve their goals (food, exits). In such a situation message's receiver can not observe message's object at the moment when message is received. It is shown, that these non-observable objects can be mapped to observable objects using local environments. Agents do not build representations of the external word, instead of reasoning on representations of the world they access the world directly through perceptions and actions and their perceptions influence their behaviour. Messages from other agents, i.e. emerging language also change their behaviour and increases effectiveness of the whole population. Language is described via two mappings: syntax (i.e. syntactic objects, words) is interpreted in semantics by the meaning mapping; the speech mapping creates for semantic objects their syntactic denotations. Words in agents language gradually become mediated semantic objects, i.e. obtain the same significance (trigger the same actions) as the actual real-word situations which they denote.