Wednesday, 5 December 2007

Robin Hunicke

I have just come back from a lecture by Robin Hunicke, a wonderful and generous speaker. If you ever get the chance to hear Hunicke speak, do so.

Hunicke has quite an impressive background. The reason that I attended Hunicke's presentation was because Hunicke's PhD research is in AI in video games.

I got a chance to speak to Hunicke briefly at the end of the presentation to get some details about Hunicke's research. Hunicke is currently researching the limitations of AI current approaches in games. Specifically Hunicke is analysing the affect of automatic difficulty adjustment by artificial intelligence in games. The hypothesis is that this approach is unsatisfactory and what is required is "consequential AI."

Consequential AI is best understood by first understanding traditional game AI which is heavily scripted thus relatively constrained, inflexible and often linear. Consequential AI in contrast is one that has a greater understanding thereby reducing the amount of scripting required. For example if you hand an AI a can of Coke, traditional AI would have a command to drink it without thought. Consequential AI would instead realise what the can was and depending on a perceived desire or need, decide to drink or not.

I think this was what Hunicke meant, I could be wrong but I do believe I got the "gist" of it. The nice thing is that I've found another person who's views align with my own research.

No comments:

Post a Comment