Joscha Bach - What does it take to build conscious Artificial Intelligence?

Artificial Intelligence is not just an engineering discipline, but also a philosophical project, aimed at the naturalization of the mind. By allowing to build testable models, AI offers a metaphysical framework and a methodology for defining and exploring mental representations, perception, agency, self modeling, attention and systemic models of psychology. At the same time, very little practical AI research is concerned with understanding consciousness and the mind. Starting from the epistemological position of computationalist functionalism, we will discuss the phenomenology of consciousness (especially second order perception and "nowness") in conjunction with hypotheses about the functionality of consciousness in a cognitive system from an AI perspective. What does a test for consciousness look like? Can we determine whether existing AI systems possess conscious phenomenology? Is consciousness functionally a learning algorithm for self organizing systems? Is consciousness simple enough to discover it via a systematic search, or will it turn out to be as complex as eg. a biological cell? Can insights from AI experiments inform our understanding of consciousness in biological systems? What are the ethical and practical implications of building conscious AI?