An Executive’s Guide to Artificial Intelligence

Share in:
Share in:

Business executives can use AI to improve their enterprise.

You have probably heard about artificial intelligence, machine learning, and deep learning. However, it is not surprising that you might not have any idea of what they really are, and how they can help your business.

If you are looking to explore user cases for these technologies for your business, then you first need to know what AI, machine learning, and deep learning are.

Artificial Intelligence

AI is when a machine is able to do things that are traditionally done only by humans, such as reasoning, perception, interaction with environment, learning, creativity, and problem solving.

For example, self-driving cars can cruise down highways without a human at the wheel. It used to be that only humans could drive. Well, not anymore.

That innovation is ushered by several factors:

  • Big data and the availability of large sets of information
  • Algorithmic advances
  • Faster and bigger computing storage and computing power

All of these are the main ingredients that make AI a reality today.

Other examples of artificial intelligence include robots, chatbots, computer vision, and machine learning.

What is machine learning?

We live at a time when it’s not enough that machines have the artificial intelligence to think for themselves. They should also learn.

How does a bunch of copper wires, circuits, and metal learn the way humans do? Machines use algorithms and artificial intelligence to detect patterns that are found in big data sets.

Looking at big data and then applying algorithms, machines are able to detect trends, make predictions, and recommend the best course of action. It processes both experiences and data on its own.

Humans do not have to program machines anymore. Artificial intelligence will tell the machine what it needs to learn. These algorithms also adapt as they encounter new data.

There are three different types of analytics that machines can do:

  • Descriptive: Tells them what happened, and is probably the most widely used form of analytics today.
  • Predictive: Tells them what is most likely to happen. This type of analytics is what businesses use to get insights.
  • Prescriptive: Tells them what to do to achieve certain business goals. This type is mostly used by top companies in every industry, as well as Internet and technology companies.

Machine learning is more focused on predictive and prescriptive analytics.

Major types of machine learning

How do machines “learn”? What are the three types of machine learning and how do you use them in your business?

1.    Supervised learning

Supervised learning is when you input data and feedback, and artificial intelligence processes them. You can liken this to a calculator wherein you input digits and the operations you want to make, and the device will calculate the numbers for you.

An algorithm will be able to find patterns and trends between the input variables and what you want to know. You train the machine to find these connections up to the time when it can be fairly accurate. You then use this “trained” algorithm on new data.

You can use this type of machine learning to predict call volumes at your call center to make sure that you have enough people answering phones.

2.    Unsupervised learning

Then you have unsupervised learning where the machine explores input data but without you telling it what the output variable should be. Unlike supervised learning, you leave the machine to figure out what to look for.

For example, you give it data on your customer profiles and that’s it. The machine will be the one to look for patterns and trends in the data and report it to you.

This is very useful when you have unstructured data or information that you cannot classify. You ask the machine to classify the data for you. Going back to our calculator example, instead of just doing the calculations for you, the machine also decides what appropriate operation to use.

You can use this type of learning for those recommenders that you have gathered. For instance, one music streaming service can recommend songs that people like you have listened to.

3.    Reinforcement learning

Think of a child who does things just to hear their parents say, “good job!”

An algorithm will perform an operation or complete a task based on motivation or rewards that it gets. This is best used when you don’t have training data to start with and you don’t know what the goal is.

There are also times when the algorithm needs to interact with the environment to learn about it.

Reinforcement learning is seen in different use cases. It can be robots that manage your inventory, or software that can optimize trading strategies.

This is also the type of learning that self-driving cars use.

Deep Learning

The last concept you need to know is deep learning.

Machine learning that involves more data and a wider range of resources is the first benchmark of deep learning.

It also needs less time for humans to process the data beforehand. Deep learning has more accurate results than ordinary machine learning. For instance, deep learning is more accurate than machine learning when it comes to image classification and facial & voice recognition.

Models of deep learning and business use cases

Convolutional neural network

This model involves a multi-layered neural network that is able to extract features from each layer to get the results you need.

This is when a machine is able to first learn about a certain image, and then apply its learning to other images. For instance, when it first sees the letter B, it will separate the vertical line, the upper curve, and the lower curve of the letter.

The machine will then be able to identify the letter B in succeeding images by identifying these components.

USE CASE: You can use images and a convolutional neural network to diagnose health diseases or weed out defective products that are in production.

Recurrent neural network

Like the CNN, recurrent neural network has multiple layers. However, it stores data in context nodes.

As a result, RNN is able to learn data sequences, rather than individual features. It can then output the desired sequences.

USE CASE: Businesses use a recurrent neural network to generate captions for images, use chatbots more effectively in handling more complex customer issues, and check whether a sales transaction is fraudulent or not.

Photo courtesy of The Cable Show.

Scroll to Top