On a windy autumn day in September 1992, I stumbled upon a special issue of Scientific American called Mind and Brain. It focussed on the biological foundation of consciousness, memory, vision, language and learning. While all the beautifully illustrated stories intrigued me, two stories really captured my imagination.
How Neural Networks Learn from Experience by Geoffrey E. Hinton, painted a picture of a network of artificial neurons that could learn to represent complicated information, where Neurons for Computers by Drew van Camp, a fellow computer programmer, outlined in great detail how such a network could be trained on a computer. As a senior student in Computing Science, this got me thrilled.
Soon I developed my own neural network that could learn to recognise hand written characters. It was both a massive success and a colossal failure: even though the network correctly learned how to recognise the characters, it was slow ... worse-than-watching-paint-dry slow.
Now, almost 30 years later, my interest in AI is stronger than ever. I consider myself to be a novice in the field, and I'm currently trying to implement Kenneth Stanley's NEAT algorithm in Swift.