The technical and scientific nature of artificial intelligence technologies make AI an intimidating topic for most people. Also, the hype around what AI can do, blurs the lines between reality and science fiction. And while you might think that AI is too “bleeding edge” for your organization, it is very likely that you are already using AI in one way or another, at work and at home. Worry no more: AI definitions are here to help.
At Novacene, our mission is to enable businesspeople without a data science background to leverage the power of AI. Some definitions of AI are helpful but lengthy. So, with this post, I will attempt to make it easy for non-technical people to understand basic concepts AI, by providing simple and to-the-point explanations.
Artificial intelligence (AI)
AI refers to the ability of computers to perform tasks that resemble humans’ abilities to perceive the world around them and to make decisions based on that perception. For example, computers have the ability to:
- read a document and perform text analysis to understand topics;
- see the objects in an image and organize them in categories; or
- hear and understand the instructions you tell it to perform.
The sensory capabilities like seeing and hearing are possible thanks to sensors like cameras and microphones. The cognitive capabilities like understanding and organizing, are possible thanks to machine learning algorithms.
Machine learning (ML)
Machine learning refers to a set of statistical methods that make it possible to predict an outcome. The larger the amount of data used to predict the outcome, the higher the likelihood that the prediction will be correct. That’s where the concept of learning in machine learning comes from. Like humans, machine learning models learn to make better decisions as their exposure to data increases.
Supervised machine learning
Supervised learning refers to the way in which a machine learning model is trained using labelled data. For example, let’s say you are trying to automate the verification of user accounts to determine whether they are real or fake. In a supervised learning scenario, you train the machine learning model using labelled data. In our example, labelled data could be an Excel spreadsheet with columns for the different attributes of an account and a column containing the corresponding label (in our case, real or fake). The machine learning algorithm would create a mathematical representation that reflects the relationships in the data, and use that representation to predict the labels for cases that it hasn’t encounter before.
Unsupervised machine learning
Unsupervised learning is another method used by computers to predict an outcomes. This method does not require labelled data, however. Instead, the algorithm uses mathematical techniques to make predictions. For example, an unsupervised learning model can be used to group user comments into different categories. In this case, the computer performs a few tasks: first, it converts the text to a mathematical representation. Then, it plots the mathematical points on a graph and groups nearby points together into clusters. These clusters represent the different categories.
In the context of AI, there is a continuum of ways in which analytics can be visualized. You’ll hear terms like descriptive analytics, diagnostic analytics, predictive analytics and prescriptive analytics. The first two refer to the ability to see the past and the present (descriptive) and the ability to understand patterns and relationships in the data (diagnostic). The last two are about predicting future outcomes (predictive) and providing recommendations on what action to take (prescriptive).
The path forward
We hope that these explanations help increase your understanding of AI. But these AI definitions are just the beginning. We are here to support your organization as it begins to plan and incorporate AI into day-to-day operations. Don’t hesitate to get in touch to discuss your needs.