Artificial Intelligence (AI), and more pragmatically, Machine Learning (ML) have been called the new electricity that’s powering the Fourth Industrial Revolution (4IR). AI is considered to be the ability for an artificial system to make cognitive decisions without human input, whereas ML is considered to be the ability for a computerized system to improve performance of an automated task over time. While AI is still futuristic, ML has been causing huge disruptions across multiple industries over the past decade. For example:
- Spam Filters – ever wonder what happened to all those Viagra ads, Nigerian prince solicitations and phishing emails that used to flood your inbox? Email filters based on ML routines have severely curtailed that flood. Google confirmed the findings in 2015, noting that machine learning reduced Gmail spam to less than 0.1% of all emails.
- Chatbots – is that a live support rep running this vendor’s web chat session, or an ML-driven chatbot? Sometimes it’s hard to tell, but even those that can tell actually prefer the chatbot to the rep.
- Recommendation Engines – Netflix’s ML-driven recommendation engine may need some tweaking since it keeps trying to get me to queue The Rock movies: watch one, get tagged for life (◔_◔). Yet Amazon did an excellent job of recommending my Xmas gift list for the family based on my historic purchases. In fact, as much as 35% of Amazon.com’s revenue has been attributed to its recommendation engine.
To learn more, download our ML Exec Guide which explains how you can avoid the pitfalls while getting started.
And the list goes on. Google, Facebook, Uber, Baidu and AirBnB are all successfully transforming their industries though the use of ML. So how can you become an ML-driven business? Here’s a simple 4-step formula to get you started:
- Create consensus and get executive buy-in.Be forewarned: for many this is the biggest hurdle, compounded by office politics, inter-team trust and resistance to change. But hey, if business transformation were easy, you’d do it every quarter.
- Prepare your data. Your Big Data efforts are finally going to pay off! …once you finish all the cleaning, labelling and staging.The time investment with data prep can be considerable. The benefit is that ML works well with messy, incomplete and noisy datasets – the exact kind of data that executive decision makers typically distrust.
- Find the right algorithm(s). Be prepared for a lot of trial and error here as you test multiple algorithms against your data.The good news is that the computing resources required to do ML are faster and cheaper than ever.
- Operationalize your models. Many organizations struggle with this last mile, which requires incorporating your ML model into a software application.To increase your chances of success, ensure your data science and software development teams can work closely together, ideally sharing the same tools and/or platforms. In this way, the process of handing off ML models will be more of a collaborative effort than a “drop and run” event as the data science team moves on to the next problem.
As you can see, transforming your organization into an ML-driven company is a complex affair. ActiveState is here to help.
To learn more, download our ML Exec Guide which explains how you can avoid the pitfalls while getting started.
