**CANCELED** [Hybrid Seminar] "AI alignment towards collective intelligence" by Prof. Ken Mogi, Sony Computer Science Laboratories, Tokyo Collective Intelligence Laboratory, The University of Tokyo
Prof. Ken Mogi
Sony Computer Science Laboratories, Tokyo
Collective Intelligence Laboratory, The University of Tokyo
Title: AI alignment towards collective intelligence
Abstract: With the rapid development of Artificial Intelligence, there is a growing interest in the alignment problem. Some argue that it is an urgent issue, in view of the possible existential risks. Here I describe some salient issues concerning alignment from cognitive and neuroscience points of view. Alignment is seen as a bridging principle linking natural and artificial cognition. I critically review assumptions about reinforcement learning, evaluation function, evolvability, and optimization. In particular, I discuss the relevance of the Goodhart's law in elucidating limits of the traditional approach and suggesting ways to make alignment more robust. I discuss relevance of alignment for collective intelligence, where correlating cognitive factors (social sensitivity, turn taking, and diversity) might be augmented and enhanced by AI alignment. I analyze relevance of organizational principles such as the Chatham House Rule. Finally, I discuss the possibility of conscious supremacy, in analogy to quantum supremacy, as a concept leading to alignment of natural and artificial intelligence, as separate from conventional computability arguments. This talk is based on the white paper and founding discussions with Takashi Ikegami (email@example.com) leading to the Collective Intelligence Lab of the University of Tokyo Komaba campus which started in April 2023.