Two new projects at OxAI Labs

09 Feb 2020

Michaelmas term was very productive for OxAI Labs. At the start of the term the team chose two topics to work on during this academic year, and spent the following couple of weeks interviewing students interested in joining the projects. We are glad to admit that there were many exceptional candidates, and we were happy to see so many Oxford students passionate about AI. The applications truly reflected a wealth of talent and passion in the community and it was a great pleasure to talk to everyone.

On the 10th of November, we held the first official meeting for all members of the two newly formed teams, and the work began. We are planning to develop the projects before the start of Trinity term. In this article, I would like to briefly describe the two research ideas, and mostly, to welcome all new members to OxAI!

Earth and Space - Satellite Image Analysis to Detect, Assess and Suggest Intervention Strategies for Wildfires

Shu Ishida (project leader), DPhil in Autonomous Intelligent Machines and Systems at Christ Church College
Dr Natalia Efremova, Teradata Research Fellow in Marketing and AI at Said Business School
Charlie Griffin, Undergraduate in Computer Science and Philosophy at Hertford College
Jan Malinowski, Undergraduate in Physics at Corpus Christi College
Louis Mahon, DPhil in Computer Science at Linacre College
Benedetta Mussati, Master’s in Computer Science at St Hugh’s College
Alexandre Szenicer, DPhil in Geophysics at Linacre College

Continuing increases in wildfire occurrences are detrimental to the ecosystem of rainforests and have direct consequences for the climate. Research shows that wildfire activity appears strongly associated with warming and earlier spring snowmelt. Additionally, forest fires release CO2 into the atmosphere, which in turn increases the rate of forest fires, creating a snowball effect.

The project proposes to apply machine learning to detect wildfire, forecast its spread, and assess its cause. The challenge will be addressed by analysing satellite images, local weather conditions, local geographic and demographic characteristics, and governmental reports and statistics on wildfire. We will provide an interactive web service to monitor the spread of wildfire and visually assess its cause and risk of future occurrences, which will be made available to the public audience. Furthermore, since more than 80 percent of the wildfires are said to result from human causes, either intentionally or unintentionally, we hope to raise public awareness by visualizing the consequences of wildfires.

AI in virtual social environments - Virtual developmental robotics in a social VR environment

Shin Ding Huang (project leader), OxAI
Guillermo Valle (project leader), DPhil in Physics at Magdalen College
Mengxi Wang, DPhil in Experimental Psychology at Christ Church College
Brandon Smith, Undergraduate in Physics at Magdalen College
James Stomber, MSc in Social Science of the Internet at St Benet’s Hall
Ryan Hughes, DPhil in Physics at Oriel College
Benjie Wang, DPhil in Computer Science at Keble College
Andy Bowery, Post Doc/Research Assistant in Oxford e-Research Centre

The aim of this project is to build a framework to train AI agents in social virtual reality (VR) environments. The main motivations are to explore how social interactions with humans can benefit the development of more human-like artificial intelligence, and whether it has the potential to pose a systematic risk to human society. This idea builds on top of research in developmental robotics, reinforcement learning, and psychology. However, research in human-AI social interaction has been limited by the use of expensive robots and sensing systems. Therefore, we propose to go beyond such limitations and lower the barrier for these types of experiments by using new VR technologies.

Our approach is to build a software framework on top of an existing social VR application (e.g. NeosVR, VRChat), and to test this framework by experimenting with agents which implement novel and state-of-the-art learning techniques such as algorithms based on intrinsic motivation and curiosity, which have shown promising results on efficiently learning complex behaviour in rich environments Oudeyer 2018.

If you are interested in hearing more about our progress or the OxAI in general, I strongly encourage you to sign up for our weekly newsletter and to visit our facebook page. Last year the OxAI Labs team were working on a model generating new levels for a VR game called Beat Saber, you can read more about it here.

Written by
Dominika Bakalarz
dominika.bakalarz@oxai.org