The multimodal turing test for realistic humanoid robots with embodied artificial intelligence
Strathearn C., Ma M.
Alan Turing developed the Turing Test as a method to determine whether artificial intelligence (AI) can deceive human interrogators into believing it is sentient by competently answering questions at a confidence rate of 30%+. However, the Turing Test is concerned with natural language processing (NLP) and neglects the significance of appearance, communication and movement. The theoretical proposition at the core of this paper: ‘can machines emulate human beings?’ is concerned with both functionality and materiality. Many scholars consider the creation of a realistic humanoid robot (RHR) that is perceptually indistinguishable from a human as the apex of humanity’s technological capabilities. Nevertheless, no comprehensive development framework exists for engineers to achieve higher modes of human emulation, and no current evaluation method is nuanced enough to detect the causal effects of the Uncanny Valley (UV) effect. The Multimodal Turing Test (MTT) provides such a methodology and offers a foundation for creating higher levels of human likeness in RHRs for enhancing human-robot interaction (HRI).