Consciousness of Crowds - The Internet as Knowledge Source of Human Conscious Behavior and Machine Self-Understanding

摘要

This position paper introduces our approach to the artificial consciousness problem, approach which we developed during our experiments with automatic common sense retrieval and affective computing for open-domain talking systems. The data about the common behaviors and usual feelings is taken from the Internet, not from a particular programmer or user. Behavior patterns are retrieved based on Schankian scripts, usual emotional consequences of particular actions are calculated for every possible outcome, and the most positive action is chosen to be performed after a confirmation. The system is not only learning the conscious behavior, it can also "know" what it is or not capable of by using simple text-mining.