Machine learning is one of those technologies that plenty of people talk about, but relatively few are putting into practice. At least, not yet. Although there are signs that may be changing.
Gartner places machine learning near the top of its Hype Cycle, next to the “peak of inflated expectations”. That’s where it’s been for a while. In fact, since at least 2015, when Google CEO Sundar Pichai said:
“Machine learning is a core, transformative way by which we’re rethinking everything we’re doing,”. Meanwhile, IDC expects a rise in machine learning spend from $19.1 billion in 2018 to $52.2 billion in 2021.
For now, most consumers may not recognise or understand the term “machine learning”. However, any internet user will be familiar with those “you may be interested this product” recommendations on e-commerce stores, or the “if you liked that film, you may like this one” message on the likes of Netflix or Spotify.
That’s basic machine learning in action. Here’s how the technology is coming to an industry near you:
Marketers’ roles are becoming more data-driven. Initially, CRM systems such as Salesforce and HubSpot paved the way. Personalising content based on users’ preferences, automating replies to keep in contact with large databases, analysing interactions for campaign monitoring.
Machine learning will take this further. Much further. Software such as Automated Insights is already writing news reports for outlets such as Associated Press. And Persado, a natural language processing algorithm, is being used by companies such as American Express, Citi, and Expedia. Instead of writing communications from scratch, machine learning will crunch the data to leave teams free to focus on editing, testing and optimising.
Manufacturing is one of the industries that has the most to gain from machine learning.
A McKinsey report (pdf) found ‘R&D cost reductions of 10 to 15% and time-to-market improvements of up to 10% are expected’ with machine learning.
Factories from the likes of Rolls Royce, Mercedes and LG all use machine learning for tasks including predicting faults before they occur, and generating more insightful data reports.
IBM is building cognitive systems to handle the growing volumes of medical data. DNA sequencing requires immense amounts of computational power. WatsonPaths will use DNA knowledge to gain insight into the best way to treat cancer.
Because, at least in the near future, machine learning is nothing like how the human brain learns. However, the key advantage of machine learning is its ability to crunch ever-larger data sets, analysing variables, at high-speed. This offers unprecedented opportunities within health, for understanding and potentially solving diseases.
Of course, as the technology evolves, this will pose new challenges for governance and compliance. One example of this concerns DeepFace, Facebook’s machine learning system.
The system has a reported 97.35% accuracy, after being trained on 4 million images uploaded by users. However, within the EU this posed data privacy concerns among regulators. After some debate, DeepFace was rolled out in 2015, but not in the EU.
The examples above show what is possible. Because here’s the thing: Machine learning doesn’t need rules for analysing data. It just takes the data, and makes its own interpretations. That’s why its potential is so great. As data volumes continue to grow, expect machine learning to become crucial to forward-thinking businesses.