By now, the race to AI — and especially to using AI in the enterprise — is officially on. But if AutoML is not in your toolbox or if you’re not at least thinking about aligning the data organization around augmented analytics, you won’t get very far.
What?
Before tackling the why, it’s important to understand what AutoML is exactly as well as what role it plays in the broader field of augmented analytics. At a very high level, AutoML is about using machine learning techniques to, well — automatically do machine learning. Or in other words, it means automating the process of applying machine learning.
Early on, AutoML was almost exclusively used for the automatic selection of the best-performing algorithms for a given task and for tuning the hyperparameters of said algorithms (in a nutshell, hyperparameters are like knobs that need to be tuned when tuning a machine learning model).
Yet AutoML can have a broader scope with later versions of auto-sklearn and tpot (and has). Its development has spurred the application of automation to the whole data-to-insights pipeline, from cleaning the data to tuning algorithms through feature selection and feature creation, even operationalization. At this larger scale, it’s no longer AutoML, but augmented analytics. Today, automated analytics can add efficiency to large swaths of the data pipeline, with the potential to impact the entire process and influence the structure of data teams long term.
Why?
Entering the AI race without AutoML or augmented analytics would be like entering a driver into a Formula 1 Grand Prix without the right car and supporting team. Sure, you can do machine learning and eventually AI without it, but it will be much slower and less efficient, which inherently means trouble scaling. Not to mention that without best-in-class tools and resources, top talent will start to look elsewhere.
So if this is the case, then the next logical question is...
How?
Getting started with AutoML and making the move toward augmented analytics might seem simple, as there are more and more tools that allow for automation of various parts of the data-to-insights pipeline. However, tooling and technology isn’t the largest problem — some major challenges to implementation are:
- Staffing and organizational: How can a company with both data scientists and analysts manage resources and responsibility in the wake of automation and augmentation? The psychology of these shifting roles is important and often overlooked.
- Transparency and trust: The more automation becomes a part of peoples’ work, the more difficult it can be to trust any results from machine learning or data projects, so transparency becomes critical here. Trust also plays a role when talking about giving more and more responsibility to citizen data scientists through increased automation.