Queen Mary University of London
I always had a fascination for automating complex tasks. As a result, my bachelor focused on software development paired with applied mathematics, and my master focused on Artificial Intelligence. I'm particularly fascinated by reinforcement learning (RL) and continual learning. I currently focus on self-motivated RL and integrating large language models with RL. I enjoy playing board games with friends and cooking during my spare time.
A description of Dominik's research:
I'm collaborating with Creative Assembly to create automated playtesting methods for Total War. I plan to achieve this goal by creating a self-motivated AI. A self-motivated AI should explore games without relying on a pre-defined goal. Instead, the AI should keep challenging itself to find new and exciting things to do in the game. One way to achieve this is through intrinsic motivation, in which we provide an AI with motivations like curiosity or empowerment. However, existing methods for intrinsic motivation usually only utilize one type of motivation. My goal is to provide agents with a wide range of motivations, hopefully creating much more diverse agents.
But, at the moment, I'm looking into enhancing agents with large language models (LLMs). The primary reason is that RL agents are slow learners and do not possess the ability to think abstractly. My hope is that the reasoning capabilities of LLMs might solve this, thus allowing me to apply intrinsic motivation on a much more abstract level than previously possible.