A Primer: Understanding Artificial Intelligence (AI)

A Primer: Understanding Artificial Intelligence (AI)

A Primer: Understanding Artificial Intelligence (AI)

Artificial intelligence is already changing the way we communicate with one another, plan our day and even shop.

Aside from these basic, consumer applications of artificial intelligence, it’s also being used by companies of all sizes today to reduce operational costs, communicate with customers and prevent cybersecurity attacks. The advantages are real, but many companies struggle to determine if it makes sense for their business.

To help you better understand the types of artificial intelligence and determine if it’s right for your business, we’ve outlined the differences and provided examples of each. We’ll cover:

What is the difference between artificial intelligence, machine learning and deep learning?

When to use deep learning versus machine learning

The difference between supervised and unsupervised machine learning

How supervised machine learning can be used

How unsupervised machine learning can be applied

What is the difference between Artificial Intelligence, Machine Learning and Deep Learning?

AI is the umbrella term used to describe the various types of artificial intelligence. Or think of it like this: artificial intelligence is the broad term that’s used to describe the idea of machines being able to carry out tasks, like humans.

Artificial intelligence is defined as a type of computer science that enables computers and software to behave intelligently. Artificial intelligence can recognize speech, learn, plan and problem solve.

Think that artificial intelligence is something that’s years away from being used in real-world business applications? It’s more pervasive than you think.

What is Machine Learning?

A subset of artificial intelligence, machine learning is the study and application of algorithms and statistical models that computers use to perform a specific task, without using explicit instructions. Machine learning utilizes patterns and inference to produce outcomes.

Let’s take a look at a few examples to get to know machine learning better.

One popular, consumer-based application of machine learning is used by Spotify, a popular music streaming service. Based on a user’s listening history, Spotify creates a playlist every Monday, known as Discover Weekly, for you to listen to. And no, it’s not put together like your favorite mixtape was in 1980 by a human.

Instead, Spotify uses machine learning to digitally create a new playlist for you of music you are likely to love. The net result; their users can’t get enough.


How is machine learning a part of this process? Spotify uses a combination of three recommendation models, according to Sophia Ciocca, who interviewed a Mikhil Tibrewal, a data engineer at the company, on the music magic behind the curtain.

When combining NLP (Natural Language Processing), collaborative filtering and audio models, Spotify manages to produce a 30-song playlist that many users love. And not a single human had to spend hours putting it together based on your listening preferences. Just think how long it would take for a human to put together individual playlists together for Spotify’s 200 million users.

Now, let’s move onto deep learning.

What is Deep Learning?

Deep learning is a subset of machine learning and many consider it to be more advanced than machine learning. Think of deep learning as machine learning, but on steroids.

Deep learning uses a network of machine learning algorithms, also known as a neural network, that is designed to mimic the human brain. When the algorithms are combined with large structured data sets, deep learning enables machines to make decisions on their own.

In comparison to deep learning, machine learning models improve their ability to predict outcomes over time, they still need human guidance. For example, if a prediction or outcome is incorrect, a human will need to make adjustments to the algorithm. With deep learning, the algorithms used can determine if the prediction is accurate on its own.

An example of deep learning involves autonomous vehicles. For example, when an autonomous vehicle is driving down the street, deep learning models help the vehicle to recognize street signs or pedestrians and in turn, make a decision about which direction or action to take.

Still confused on the difference between machine and deep learning? Let’s break it down again.

Machine learning uses algorithms and data to make predictions. However, machine learning does not enable the machine to learn whether it’s predicted outcomes are correct or not. And if they aren’t correct, the machine doesn’t know how to fix it without a human intervening.

Deep learning utilizes multiple, layered algorithms that form into a neural network, much like the neural network in the human brain, that can learn and make intelligent decisions on its own. Deep learning is more intelligent than machine learning in that it can help machines learn whether a prediction is correct and thus, in turn, readdress the problem.

When to Use Deep Learning vs. Machine Learning

Not sure when to use deep learning versus machine learning? You’re not alone. There are two things you’ll need to apply deep learning versus machine learning. You’ll need data (a lot of it) for your machine to learn from and substantial computing power. If you’re without these two things, you’ll have to use machine learning.

However, as challenging as deep learning is, when it’s executed properly, the results are impressive. Let’s review a few examples of deep learning to better understand how it works.

One of the most recognized applications of deep learning is chatbots. Chatbots are often used to provide sales assistance or customer support, the most common being the latter. These chatbots use workflows and deep learning to answer customer queries. As the service is used more and more and the machine gathers more data, deep learning enables a near-human conversation.

Another example is the well-known autonomous vehicle. In this case, deep learning models train themselves to recognize street signs while another may be trained to recognize pedestrians. Informed by millions of AI models, a car driving down the road is able to react in time in order to respond in time when in actively driving in traffic.

Now, let’s jump into the differences between supervised and unsupervised learning.

The Difference Between Supervised and Unsupervised Learning

Think of supervised and unsupervised learning as the types of tasks you’ll ask your machines to learn. To help you understand the difference between the two, let’s first define each one.

What is Supervised Learning?

Supervised learning is defined as the machine learning task of learning a function that maps an input to an output based on example input-output pairs. Or think of it like this: supervised learning uses artificial intelligence to build models that map data to the correct answer. In other words, you need both training data and the expected answers.

What is Unsupervised Learning?

Unsupervised learning occurs when you have input data that has no labels, or correct answers. Unsupervised machine learning algorithms infer patterns or outcomes without reference to known, labeled outcomes. It’s an approach during which machine learning enables the software to learn from the data without giving it the correct answers.

How Supervised Machine Learning Can Be Used

Let’s examine a real-world application of supervised machine learning, as provided by Simplicable. Let’s say that a robot is learning to remove and separate recyclables from trash. As the conveyor belt rolls along, the sorter places recyclables into bins and labels each with an identification number.

Then, once a day, a human examines the bins and informs the robot which items were correctly sorted. In turn, the robot uses this information to improve its future decision making. This is an example of supervised machine learning since a human is required to help the machine learn how to produce the desired outcome.

How Unsupervised Machine Learning Can Be Applied

Let’s use the aforementioned example to demonstrate how unsupervised machine learning can be applied.

In the previous example, a human would not be available to tell the machine whether it successfully removed the recyclables from the trash. The machine would have to learn that on its own.

In other words, the machine learning model needs no supervision. Instead, it works on its own to discover data patterns and insights. Unsupervised machine learning enables more complex processing tasks than supervised learning like clustering and association.

While you need to do the due diligence to determine if AI makes sense for your business, the applications and benefits are real today. If you’d like to learn how ONE Tech applies machine learning for the benefit of industrial IoT, learn more here.