Last time we covered the definition of AI. You’ll probably have heard the terms AI and ML being used interchangeably in the mainstream press or at events, but they are actually quite different topics. Machine learning is a sub-category of AI.
"Machine learning is the scientific field concerned with algorithms and statistical models which allow computers to complete a specific task without explicit instructions, relying on patterns and inference instead."
To carry on with the chat bot example from earlier, the ML version is called Natural Language Processing (NLP). Whereas the above chat bot used a decision tree to pick the correct response to a question from a set of pre-written answers; NLP applies algorithms to identify and extract natural language rules from data of an entire language and then computes them in a more conversational way. Alexa is a perfect example of NLP; she is able to understand the underlying patterns that structure the English language to understand and respond to requests in a fluid way. She requires a lot more data of conversations than a simple chat bot, who only requires a script of questions and answers.
Machine learning goes a level of complexity deeper than AI when referring to types of AI systems: advancing from pure logic (if input x, then output y in conditions z and w) to the ‘understanding’ and inference of patterns in data (breaking a sentence down into syntax and semantics categories to allow a computer to ‘understand’ its meaning).
Deep learning, in turn, is a sub-category of ML. Specifically, it is a class of machine learning algorithm. It is one of the most complex forms of computational information processing as all deep learning systems mimic the way neurons in the human brain help to process complex information. Almost all deep learning applications are based on artificial neural networks. There are many different types of neural networks. Two of the most common that are brought up in conversation are Convolutional Neural Networks (CNNs) and Adversarial Neural Networks (ANNs).
The term ‘deep’ in ‘deep learning’ refers to the number of layers through which data is transformed.
Deep learning is most commonly used in advanced natural language processing and image recognition (for example, the military tank identification AI mentioned earlier, which is based on a real application in the US Military in the 1950s whereby the AI kept recognizing the time of day in images instead of tanks, causing very poor accuracy results in an early experiment of training an artificial neural network).
For more information on Convolutional Neural Networks (CNNs) check out this video here For more information on Adversarial Neural Networks (ANNs), also known as Generative Adversarial Networks (GANs) check out this video here