Machine Learning, AI, Neural Networks and Cognitive Computing: A Primer for CIOs
The exponential growth of digital data in the past decade, combined with groundbreaking advances in intelligent computing, offer enterprises unprecedented opportunities to solve business problems, gain competitive insights and advance market and product research.
For CIOs who are in charge of the success of their enterprise’s intelligent computing initiatives, choosing the right technologies and tools can be daunting, as the various terminologies in the world of intelligent computing are confusingly similar. For instance, what is machine learning? Is it the same as artificial intelligence (AI), or does one enable the other? Where do deep neural networks and cognitive computing fit into the mix? It’s enough to make a CIO’s head spin.
Though some of these terms may overlap, there are important distinctions between them. The following are several key definitions that will help CIOs better understand the intelligent computing landscape:
The term “artificial intelligence” was first coined in 1956, and though definitions may vary, one from Stanford University is a good place to start: “It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence.”
Put another way, AI is all about making computers think and learn like humans. As Lynne Parker, director of Information and Intelligent Systems at the National Science Foundation, told Computerworld, AI is “a broad set of methods, algorithms and technologies that make software ‘smart’ in a way that may seem human-like to an outside observer.”
The definition of machine learning depends on whom you ask, but the only real debate is whether it is synonymous with AI or a subset of it. For example, machine learning researcher and Sensai Corporation CTO Monica Anderson calls machine learning “the only kind of AI there is” on Quora, while TubeMogul Chief Scientist John Trenkle argues on Video Ad News that AI is a “larger entity that actually utilizes the outcome of machine learning.”
However you categorize it, machine learning is a program using data as feedback to improve performance. Among other applications, machine learning can be used to filter spam, categorize content, detect human faces or segment customers. Parker told Computerworld that the term “machine learning” tends to be more common in Europe, while AI is the preferred term in the U.S.
Deep Neural Networks
Robert Hecht-Nielsen, Ph.D., who invented one of the world’s first neurocomputers, defines neural networks as “computing system(s) made up of a number of simple, highly interconnected processing elements, which process information by their dynamic state response to external inputs” on the University of Wisconsin-Madison website.
The premise behind neural networks is that they are patterned after the neuronal structure of human brains, though on a less complex scale (for now). Neural networks feature layers of connected nodes that can trigger reactions to data, and deep neural networks have more layers that can be trained.
With cognitive computing, systems use natural language processing, data mining and pattern recognition to learn in ways similar to the human brain. IBM’s Watson is probably the best-known example of a cognitive system.
Cognitive computing can be used to help health care organizations provide better care to patients with chronic conditions, to enable retailers to better engage customers or to simply vanquish humans in “Jeopardy.” Cognitive computing is also being harnessed across a number of industries in mobile applications and the Internet of Things.
These basic definitions should help CIOs determine which advanced technologies will be most beneficial to add to their enterprise infrastructure.