A Conceptually Different Approach to the Empirical Test of Alan Turing

摘要

The test for general human-level intelligence proposed by Alan Turing at the dawn of computers remains to be the best-known ultimate challenge for Artificial Intelligence today, if understood more generally than the originally proposed paradigm, despite all its controversies and weaknesses. Yet, a better-posed challenge is needed in the era of new technological capabilities and ambitions: a challenge that can drive research toward practically vital achievements. In this work, a new general challenge based on a Turing-test-like paradigm is formulated, together with a roadmap based on it. It is argued that the proposed roadmap is vital for the integration of future robots and virtual agents into the human society.