KISS
As my degree was in Cognitive Science rather than Computer Science, pretty much all we did was study NP hard problems and non-deterministic neural networks. The up side of this is that anything that I can recognise as being deterministic, I know I can solve using basic (or not so basic) algorithms, by breaking the problem down using various problem-solving techniques. It's surprising how many seemingly intractable problems resolve down to using a combination of one or two techniques and a couple of algorithms. The down-side of my degree is that I find very little software in the commercial world to be particularly challenging, just tediously long-winded. I long to work on neural-network designs but so far little has escaped from AI research labs to be used in real-world applications. Even visual processing tends to rely on various mathematical techniques involving massive number crunching. The closest we get is using Hidden Markov Models to fudge the parts where deterministic algorithms fail.
Biting the hand that feeds IT
