SAS recently produced a helpful video to cut through the misunderstandings that have developed in the market concerning the distinctions between AI, machine learning, and deep learning (many marketers using the terms interchangeably)
In summary:
> AI is a term for a solution which often utilizes machine learning as a component method to emulate human decision making
> Machine learning is not AI, rather it represents a broad set of methods which often support AI solutions
> Machine learning encompasses a large set of algorithmic statistical techniques for pattern extraction and decision making from data
> Deep learning is a family of machine learning techniques which support learning from data representations, deep neural networks being a popular example
> Deep learning has produced efficacious results especially in the areas of image and speech recognition, hence its frequent use as a component in AI solutions
> Deep learning is not always the best machine learning technique – a champion-challenger approach should be applied to test many ML techniques
> Deep learning techniques do not absolve implementers from the need to treat and understand the data they feed into the algorithm – garbage in-garbage out
> Machine learning correlation does not equal causation: ignoring explanatory theory when developing and testing algorithms often leads to overfitting and misunderstandings between what the implementer intends versus what the algorithm produces – see Science ‘The Parable of Google Flu Trends’
http://www.few.vu.nl/~acs550/The%20Parable%20of%20Google%20Flu%20(WP-Final).pdf
