Neural Networks are developed using learning algorithms that mimic the human neural network structure. By examining how neural networks are trained, we can start to understand how they can produce a variety of patterns.
For example, by analyzing every aspect of a photo of your face, a neural network might be able to determine that you’re male or female based on how your eyes, mouth and hairline are positioned.
They are applied to problems in many fields such as image recognition, speech recognition and machine translation. When we apply these techniques to natural language understanding, we tend to treat them as an emergent phenomenon.
A fascinating paper by Marvin Minsky who is an American cognitive scientist concerned largely with research of artificial intelligence at MIT in the year 1970 presented one of the first theoretical analysis of this problem. He showed that a well-designed network of neural connections can make a good approximation to a human brain, but the networks are prone to a certain kind of systematic error that’s both natural and difficult to avoid.
Admittedly, deep learning is still in its infancy, and it takes years of training for a neural network to become a capable machine learning tool. This is why you’ve probably seen it in research papers or videos where data is fed into it, and a certain desired output is generated out of it. Deep learning has the potential to be extremely useful in nearly any field that you could think of.
Since not everyone can afford the expensive neural network training and evaluation machines, softwares are used to teach the machines to recognize objects and other people from pictures. Google has built an artificial intelligence called the DeepDream Generator by engineer Alexander Mordvintsev which uses a convolutional neural network to find and enhance patterns in images via algorithmic pareidolia, thus creating a dream-like hallucinogenic appearance in the deliberately over-processed images. The “Dreamer” computer is almost indistinguishable from a human in terms of cognitive ability, and it can take any video picture and generate captions or make an image of itself.
Fascinating isn’t it?