An interesting (and extremely useful) character turns up

28 November, 2012

 - credit:

Dr Joel Pitt is at Dragonfly for the next few months, doing contract work developing websites and back-end systems. Joel describes himself as part scientist, part software engineer, and fascinated with artificial intelligence and consciousness.

“I’ve been working in a business setting for the last year, so it’s great to be able to just focus on code again,” he says.

Joel has a PhD in eco-informatics from Lincoln University. His thesis involved building models to predict the dispersal of an introduced invasive species based on climate, available habitat and geography.

“I took the idealised mathematical models based on data collection that others had developed, and put them into a GIS, so they could be applied to real environments. The end goal was to create a modular system for the simulations that would be applicable to a range of invasive species, such as Argentine ants, Buddleia, Scotch broom and old man’s beard.

“Traditionally, people generally looked at the eventual distribution of an invasive species, but my research considered the progression from invasion to a potential range, and how quickly it would happen. If you know that, you can make an informed decision about whether to intervene or not.

“If a species is going to progress rapidly, there’s probably no point in trying to stop it unless it’s detected very early, but if you know it will be a slow spread, then it’s worth investing in a treatment. Also, if you know how geography will shape the spread, then you can try and optimise where you put monitoring and treatment.

AgResearch and CSIRO are currently using Joel’s model to evaluate the long-term efficacy and efficiency of different weed management strategies.

After his PhD, Joel moved into AI research, working on OpenCog, an artificial intelligence project using open-source software.

“We were trying to create an autonomous learning system that could control robots or other agents in virtual worlds. It brought together lots of traditional machine learning (a sub-discipline of AI) methods into one framework and was therefore quite powerful.”

That experience took Joel to the Hong Kong Polytechnic University in 2011 to lead a research team using OpenCog. The goal was to build a virtual learning environment where a character would learn how to interact with the environment without having any prior knowledge of it. The research has application in gaming, allowing developers to give their characters more autonomy.

Joel later joined a Hong Kong based startup, Demand Analytics, as their chief technical officer. At the beginning of 2012, he helped the company set up a small development office in Wellington.

Edward was introduced to Joel through a mutual friend some years ago. “Joel is a good fit for us, with his science background, but also strong computer science, machine learning and programming experience. It is great to have him in the office.”