NWO AiNed Fellowship grant awarded to Filip Ilievski

Human-centric AI aims to enhance the unique abilities of humans with the processing power of AI models. While humans excel in empathy and consideration for others, AI models can process vast datasets and optimize repetitive tasks. This opens the door to developing AI assistants that can augment human capabilities. However, despite the strong performance of state-of-the-art AI methods on benchmarks, the broad application of AI is limited by a lack of common sense. Effective collaboration between humans and AI requires more than just mimicking human language abilities; it necessitates a deep understanding of psychology, naive physics, cultural values, and the integration of language and vision.

Filip Ilievski's project will focus on advancing our knowledge on how human-centric AI agents with common sense can perform reliable mental modeling in human-robot conversations, verify their reasoning, and reason about complex situations such as those in videos. These AI agents will learn to understand the desires, goals, values, and intentions of humans, using this knowledge to anticipate human behavior. This includes understanding cultural values, ethical and moral aspects of interaction, and how signals from multiple modalities converge.

The project promises the development of AI systems that can not only simulate human language but also have a deeper, common sense understanding of the complex world we live in. This development will build further on advances in knowledge representation, neural language and vision models, and human-AI collaboration.