It’s a great time to start thinking about using Machine Learning. The technology is powerful mature and accessible. App developers are embedding the technology in their solutions, but you can also build custom solutions.

While there are plenty of expensive failed AI & ML projects, you can greatly reduce the risk by following some steps.

This post documents some takeaways from a recent presentation I gave on ML.

machine learning

What is Artificial Intelligence?

Intelligence can be described as the ability to acquire and apply knowledge and skills, then using this knowledge to adapt.

AI is intelligence demonstrated by machines. Performing tasks normally requiring human intelligence.

AI is also a very broad term that can be used to describe all associated fields & sub-fields of AI. Marketing departments have leveraged this by promoting any product loosely associated with AI “Now includes AI!”

Why Machine Learning?

Machine Learning (ML) is a sub field of AI using existing data to predict future behaviors, outcomes, and trends. ML gives computers the ability to learn without being explicitly programmed.

This is the technology that is currently most accessible to business. There are many cheap or free ML services, plus it reduces the need for expensive software development.

Why should you care?

If Machine Learning allows you to predict the future, then it’s definitely of interest to all companies. Who doesn’t want some help to better understand your products, customers, markets and more.

What can it do?

ML is broadly split into 2 main areas:

  • Supervised: Manually train the ML algorithm on existing data. This builds a “model” which allows predicting an outcome with new/unseen data.
  • Unsupervised: Give an algorithm a lot of data and let it try to find patterns and relationships within it.

Some examples include:

  • Calculating risk - for example in insurance & finance industries
  • Personalization - deliver recommendations to users (eg. Netflix & Amazon recommendations)
  • Customer satisfaction prediction based on user activity
  • In healthcare, allow early identification of patients
  • Image analysis - recognise object & people in images, text analysis
  • IT Security & spam email filtering

See part 2 for some practical advice