Affect-based Planning for a Meta-Cognitive Robot Sculptor: First Steps

Selmer Bringsjord, John Slowik, Naveen Sundar Govindarajulu, Michael Giancola, James T. Oswald, Rikhiya Ghosh

Abstract: Today’s so-called “generative AI” is in the minds of some capable of specifically generating genuine visual art; DALL-E is an example. While in fact intensely skeptical, we grant for the sake of argument that the likes of DALL-E can in fact generate genuine visual art. However, we observe that fine visual artistry is by no means monolithic; in particular, some forms of fine visual artistry are seemingly harder than others for artificial agents to achieve. In the traditional ontology, perhaps the very hardest type of fine visual artistry for an AI to achieve is literary sculpture, an activity carried out in the human sphere at the very highest level by Rodin. Artificial agents that operate in this sphere must for obvious reasons be cognitive robots. We explain such literary sculpture in broad terms, making clear that it’s undertaken by the sculptor in question with the aim of bringing about certain affective states in the mind of the viewer of the sculpture. In short, literary sculpting is affect-driven. We provide a case study of robot sculpting, with help from pre-existing logic-based formalisms and automated-reasoning technology, and the cognitive robot PERI.2, operating as a sculptor. To our knowledge, ours is the very first foray into literary sculpture in AI and cognitive robotics.