John Levine

Timeline

1. 1992–1994

   The IDAS documentation project

   At the University of Edinburgh, he works with Ehud Reiter and Chris Mellish on IDAS, a project that uses natural-language generation to produce technical documentation from a domain knowledge base. The three present it in March 1992 at the Conference on Applied Natural Language Processing in Trento and write up its lessons in 1994 for the journal Applied Artificial Intelligence.

2. 2001

   Adversarial planning for Go

   With Steven Willmott, Julian Richardson and Alan Bundy, he publishes "Applying adversarial planning techniques to Go" in Theoretical Computer Science in 2001, using hierarchical task network planning for goal-directed play. He also chairs PLANSIG 2001, a conference on AI planning and scheduling systems.

3. 2002–2003

   Research fellow at Edinburgh

   In May 2002, he speaks at the Informatics Jamboree on ant colony optimisation and aggressive local search for bin-packing and cutting-stock problems. By 2003, he is an Informatics Research Fellow at the University of Edinburgh, applying evolutionary search and learning to AI planning problems; he holds an MA, an MPhil in Computer Speech and Language Processing and a PhD in Computer Science, all from Cambridge.

4. 2006–2010

   Evolving knowledge for AI planners

   At Strathclyde, he is principal investigator, with Maria Fox and Derek Long, of a 2006–2010 project that uses evolutionary learning to acquire control knowledge for AI planning. In 2007, he, M.A. Hakim Newton, Fox and Long present an ICAPS paper in which a genetic algorithm learns macro-actions for arbitrary planners and domains.

5. 2013–2014

   General video game playing

   The 2013 Dagstuhl chapter "General Video Game Playing", which he writes as first author with Simon M. Lucas, Risto Miikkulainen, Tom Schaul, Tommy Thompson and others, proposes a competition for agents that play video games they have not seen before, along with a Video Game Description Language. In 2014, he and Philip Rodgers test Monte-Carlo tree search and averaged depth-limited search on the game 2048.

6. 2018

   Deceptive games for AI agents

   Working with Damien Anderson, Matthew Stephenson, Julian Togelius, Christoph Salge and Jochen Renz, he coauthors "Deceptive Games" in 2018, a set of VGDL games built around specific cognitive biases and run in the GVGAI framework. Each tested competition agent proves vulnerable to several kinds of deception.

7. November 2021

   Director of undergraduate teaching

   In November 2021, as a Senior Lecturer, Director of Undergraduate Teaching in Strathclyde's CIS department and a Fellow of the RSA, he gives a talk arguing that computer science belongs in schools as a fourth science subject.

8. 2023–2024

   Senior teaching fellow at Strathclyde

   His 2023 study with Francesco Meggetto, Crawford Revie and Yashar Moshfeghi predicts music skips using deep reinforcement learning. By 2024, he is a Senior Teaching Fellow in Computer and Information Sciences at the University of Strathclyde, and he coauthors a 2024 paper on automated marking in undergraduate programming classes with Martin Goodfellow, Andrew Abel and Konstantinos Liaskos.