[Nusselt]

I still have close friends from uni involved in some aspects related to AI (they've even got me into a rodney brooks lecturer), specifically machine vision and vision recognition (if we're talking in terms of current matrix multiplication type mathematical tools). Even the name 'neural network' to me is a misnomer, there may be a level of 'learning' involved but that doesn't mean 'understanding', all neural networks effectively does is take a matrix of data and store it in a way that can be recovered, tying up more than one layer of neural network or increasing the processing power that it runs on won't alter the fact that a computer is in no way reaching any real level of 'life' (which if we're honest is the aim here). As nodrog said pattern recognition has been achieved in very tightly controlled conditions within industrial machine vision but trying to take those tools to other everyday applications has prooved near on impossible (even with fuzzy logic). The only real success i can think of has been number plate recognition (and even that required standardised number plates), theres shit loads of research going on to try and develop AI related software for security operations (facial recognition of terrorists/criminals) but they aren't really getting anywhere, these kinds of applications relative to what they're trying to achieve (an artificial intelligence) are pretty insignificant, if the tools they use are more or less based on the same principles, i don't see how they can achieve the latter if they can't even figure out the former.