Leave No Trace

The public lecture at Wolverhampton University last night, on ‘Building Intelligence:  Autonomous Characters in Virtual Environments’ wasn’t quite what I expected. I had interpreted the phrase “interactive multimedia applications” to suggest a wide variety of contexts, possibly including art installations. The lecture, however, after listing different types of environments and purposes, focused entirely on games. That isn’t necessarily a bad thing, though, and The Guardian coincidentally points out today that games tend to be overlooked by critics.

The University claims to be researching artificial intelligence, but something troubled me about the descriptions used in the lecture. It was aimed at a non-specialist audience, so I assume that they do far more than was described, but it seemed to me that the examples given were of merely evaluating options from pre-defined lists until a successful outcome was found.

I know that the definition of intelligence in this context is debatable, but I had hoped to hear some discussion of systems reaching logical conclusions from basic precepts, even if that’s still a distant goal. At the end, I asked if they’d had any experience of unexpectedly complex behaviour spontaneously occuring, but all he responded with was a brief description of the concept of emergent behaviour and mentioned flocking.

I’m no expert in the subject, but there seems to be potential in the cross-over between genetic algorithms and neural networks to allow systems to experiment at random and learn from the results, which could then inform future experiments. I’d like to explore this further, but at Wolverhampton they write a lot of code in C++, which I wouldn’t dream of learning. Life’s too short for that.