Steve James

PhD candidate and Associate Lecturer at the University of the Witwatersrand


My current research interests lie in learning symbols for planning in continuous, low-level state spaces. In particular, I am interested in how best to learn symbolic representations of environments which can be transferred to other domains that an agent may face. This work is supervised by Benjamin Rosman of the CSIR, and George Konidaris at Brown University.

Previous work focused on developing a better understanding of Monte Carlo Tree Search and the effect of different simulation policies on the overall performance of agents in environments with varying characteristics. The results of this work can be seen on my Google Scholar profile here.


I am a PhD student at the University of the Witwatersrand in Johannesburg, South Africa, where I've spent my entire academic career. My interest in AI stems from writing my own chess engine - it should therefore come as no surprise that I am very interested in games, be they board, video or logic. I am also a keen programmer, having worked as a Java and Android developer, and have developed web apps and sites (such as this one).

Away from this, I am a massive sports geek, metalcore fan and professional sleeper.