Alan Turing was a pioneer in the concept of machine learning. For those of you who have seen “The Imitation Game,” you got a glimpse of an early prototype that had moderate success. (If you haven’t seen the movie, I highly recommend it.) Early computers only did what you told them to—albeit much faster than a human could. Turing felt that for a computer to truly be successful, it should be able to deceive a human that is interfacing with it. To this end, he created a test in 1950—the infamous Turing test. And a machine achieved moderate success in 2014 when a chatter bot (pretending to be a 13-year-old boy called Eugene) was able to fool a third of the judges in its Turing test.
Although computers are not fully conversational yet, they have progressed tremendously over the years. Text analytics and voice recognition were part of some of the early improvements. There are also many tools to help determine sentiment analysis. Many of the future advances will come from the areas of deep learning and artificial intelligence. The subset of artificial intelligence that has seen greater acceptance in the last decade is machine learning. Fortune Magazine claims that 81% of CEOs surveyed believe that machine learning is key to the success of their company.
By definition, machine learning is any time new data can be analyzed and the system can apply intelligence without human intervention. Many solutions have deployed that concept by embedding simple algorithms in the application to generate a forecast or create a propensity score of some type of behavior. This is a tremendous improvement over the traditional advanced analytic process that gave impressive results but required substantial manual intervention.
Moving to Automated Machine Learning
Consumer expectations of immediate engagement with a personalized response, whether for sales or support, has minimalized the effectiveness of traditional methods. This need has driven the birth of automated machine learning. As stated previously, machine learning is embedding any intelligence into an application or work flow. Once that algorithm has been created, it can’t help but grow old or stale and lose its effectiveness.
Implementing automated machine learning requires three phases:
- Automated model creation,
- Automated end-to-end production, and
- Automated integration into the application.
Automated machine learning allows the system to determine the most effective models. Automation also increases accuracy by retraining in-database regularly and scores new data as it’s ingested. Implementing these phases will provide consumers with the right offer to the right person at the right time.
Companies that implement all three phases of automated machine learning will have a competitive advantage over the rest of the pack.