Uncommon sense: a systemic approach to beliefs of AI agents

When discussing the topic of game AI, the main focus is typically on decision making algorithms and – in some genres – perception simulation.

But between those layers lies a whole world of belief building: storing historical observations, querying agents’ memories and inference of new information.

The talk attempts to shed some light on this underexplored territory, by describing a domain-agnostic, systemic approach to belief building, designed to operate in a wide range of game environments.

BIO:

Head of AI Development at QED Software. Responsible for development of Grail AI framework. Rafał has 7+ years of experience in the gamedev industry and is
skilled in traffic simulation in mobile games. He has also worked on AI for tactical & strategic games, in-house and for third-parties (Tactical Troops: Anthracite Shift, Ancestors Legacy).