Artificial Intelligence or Machine Learning?

I like to believe that after 25 years or more of playing with technology I have a respectable tech karma quotient. I respect technology and because we are talking about technology, it feels nothing for me, or does it?

Everyone must remember at least one science fiction film they have seen where computers take over the planet, I can name half a dozen without even thinking. Artificial intelligence and machine learning are not about computers taking over the world though are they... This notion of sentience is most likely something we need to worry about in the next century, not the next decade. After stating that, I can hear people shouting at their screens already; “but this person from this project claims this fact regarding machines and sentience - artificial intelligence has rights too you know!” The reality is we are not talking about “Ghost in a Shell” type engineering here, we are talking about servers with big graphics cards crunching numbers – There is a difference, just like there is a difference between artificial intelligence, machine learning and just a computer crunching some numbers.

Artificial intelligence and machine learning are closely related but distinct. Machine learning is fundamentally a subset of artificial intelligence, which is an overly generic term used to describe hardware or software that enables a machine to imitate human intelligence.

Machine learning is just one technique to achieve that intelligence by utilising algorithms to collect and analyse data and “learn” from that data – It is then able to make supposedly intelligent decisions. There are many other ways to deliver artificial intelligence including computer vision and natural language processing.

Since machine learning is a common technique for delivering artificial intelligence, most organisations adopting artificial intelligence solutions will inevitably end up implementing machine learning. This can be seen by the artificial intelligence in today’s smartphones which is delivered by using machine learning for features like predictive text, speech recognition, face unlock, and personal assistants.

Machine learning is about extracting knowledge from the data. It enables a system to make predictions or decisions using historical data without being explicitly programmed to do so. The process uses a massive amount of structured data so that a model can generate accurate results or provide predictions based on that data.

It works via an algorithm that learns on its own using historical data. It works only for specific domains. For example, if we are creating a model to detect pictures of specific animals, it will only give results for such images, and discard other animals. As discussed machine learning is being used in various places such as for online recommender systems, Google search, Email spam filters et cetera.

A good example of artificial intelligence is robotics. Modern industrial robots have the ability to monitor the accuracy and performance of their own work and detect when maintenance is required to avoid expensive periods offline.  

In general, artificial intelligence solves tasks that require human intelligence while machine learning is a subset of artificial intelligence that solves specific tasks by learning from data and therefore making predictions about the future.

By definition, this means that all machine learning is artificial intelligence, but not all artificial intelligence is necessarily machine learning.