The State of Analysis

Currently machine learning is the torch-bearer of data analysis, and cognitive computing and true artificial intelligence appear to be (somewhere) on the horizon, but most data analysis, including machine learning, has evolved directly from its roots in statistics.

The result is that analytics is on a trajectory towards increasingly intelligent machines using some version of the same statistical models that have been around for decades or longer.

Counter to this process of iterative improvement, we are proposing Autonomous Machine Learning as a class of solutions that exists between the analytics tools that require extensive human guidance (statistics and ML) and true cognition (AI).

A New Solution in Machine Learning

Iteration on existing methods is important, but the legacy of statistics has limited our ability to make real, transformative changes to the way that humans go from information (data) to knowledge (rules) to action (prediction triggered events). With autonomous analytics we must be able to create more than rough approximations, but rather the prescriptive rules that enable seamless automation.

This new distinction is needed because it emphasizes the reduction of human input in data analysis and the hardware and software requirements needed for actual cognition. Today’s machine learning tools are capable of finding relationships in data, but lack the literal intelligence to process and understand language or truly learn concepts and facts. Autonomous machine learning algorithms will bridge that gap by reducing large amounts of data to the key patterns that define analyzed data sets, moving analysis beyond the paradigm of human data exploration.

Not only will autonomous machine learning tools become integral to the data economy, but these advances will become necessary to realize commercially viable cognitive computing.