For much of the past decade, Machine Learning, Predictive Modeling and Artificial Intelligence have only been accessible to most advanced organizations, like Facebook, Microsoft and Google. These organizations have completely disrupted the technology industry with their innovations and investments. Earlier this year, Forbes (link) reported a 6x increase in startup investments and 14x increase in number of startups related to Artificial Intelligence since 2000, as well as exponential growth in revenue generated by Artificial Intelligence applications over the next 7 years. These numbers have undoubtedly increased since then.
So, how can we access these capabilities without waiting years to build a team of Data Scientists?
Automated Machine Learning has matured dramatically over the past few years in parallel with the computing power of the public and private cloud providers. Whether the goal is to identify valuable product offerings, maximize brand awareness or minimize supply chain costs, the ability to leverage massive banks of CPUs and GPUs has created the possibility of training millions, billions or even trillions of different predictive models in search of the one that provides the most value. This brute force approach was impossible as recently as a few years ago; but the exponentially increasing performance of the modern cloud is shattering these barriers every day.
Many Automated Machine Learning vendors are also introducing best practices and techniques from crowd-sourcing platforms, such as Kaggle, to combine the overwhelming computing power of the cloud with the expertise of the world’s brightest Data Scientists. Some of these tools are capable of automatically performing data visualization, data cleansing, feature engineering, model diagnostics and model evaluation against data sets containing millions or billions of rows of data, effectively automating the entire machine learning process. The important question becomes, “With this level of automation, how do we guarantee value, quality and the human element?”
As with any other technological advancement, Automated Machine Learning is only as valuable as the teams that leverage it. In order to truly show value with an organization, there must be investment from leadership into ensuring that data quality, data management and data lineage are upheld throughout the entire process. In addition, care must be taken to ensure that the results of the automated processes are valuable to the business and capable of solving real-world problems with all of the associate nuance. This means that Automated Machine Learning is not a replacement for modern analytics teams, but a tool for augmenting their productivity, allowing experts to invest their time in identifying, clarifying and solving hard problems, while the machine handles the grunt work.