Data Demystified: The Difference Between Data Science, Machine Learning, Deep Learning, and Artificial Intelligence

Welcome to the third part of our month-long data demystified series. As part of Data Literacy Month, this series will clarify key concepts from the world of data, answer the questions you may be too afraid to ask, and have fun along the way. If you want to start at the beginning, read our first entry in the series: What is a Dataset?

In this entry, we’ll explore the differences between various subfields within data you may hear about daily. The blog post will outline the difference between data science, machine learning, deep learning, and artificial intelligence.

Unpacking Data Jargon

Data science, machine learning, deep learning, and artificial intelligence are terms that have crept into mainstream conversations in the last few years. Still, these terms are frequently used without a proper explanation of what they mean. While there isn't an official definition for any of these terms, and while experts argue over the exact boundaries, there is a growing consensus on the broad scope of each term. Here’s a breakdown of how these terms can be defined:

• Artificial Intelligence are computer programs that can behave intelligently, reason, and learn like humans. 
• Machine Learning is a subset of artificial intelligence focused on developing algorithms with the ability to learn without explicitly being programmed. 
• Deep learning is a subset of machine learning. It is responsible for many of the awe-inspiring news stories about AI in the news (e.g., self-driving cars). Deep learning algorithms are inspired by the brain's structure and work exceptionally well with unstructured data such as images, videos, or text. 
• Data science is a cross-disciplinary field that uses all of the above, amongst other skills like data analysis, statistics, data visualization, and more, to get insight from data.

Artificial Intelligence: Programs that Behave Intelligently 

Most software includes logic written by programmers to specify precisely how it should behave given some user input. For example, on a shopping website, if you click "Add to cart," there will be code to add the item to a cart object and code to update what is displayed on the website. The website doesn't need to think; it just follows the precise instructions in its software.

By contrast, with a customer service chatbot, users can type in so many varied prompts that it would be impossible (or at least not economically viable) to program explicit instructions on how the chatbot should respond. Instead, the chatbot must "think" to understand what is being asked to come up with a response. 

While a chatbot system isn't intelligent like a human, it is powered by a machine learning model. An algorithm trained using text data can understand a specific problem (more on that later!). The chatbot performs a task without a human telling it exactly how to decide what to say in response to the message it receives, so it can be considered artificially intelligent. 

Chatbots are an example of a machine learning-based AI system. There are also examples of AI systems that are not based on machine learning. Typically these would be called expert systems, which are essentially a series of if statements. An example could be medical diagnosis software that asks patients multiple choice questions about their symptoms and provides potential diagnoses based on their inputs.

Understanding and responding to text is a typical application of artificial intelligence. A similar idea is being able to respond to speech, as seen by computer-based personal assistants like Alexa, Siri, and Google Assistant. Understanding images or video also requires artificial intelligence. This has many applications, from security cameras detecting objects to search engines displaying the right images to self-driving cars.

Machine Learning: Algorithms that Learn from Data

Since artificial intelligence makes use of machine learning models, many people consider machine learning to be a subset of artificial intelligence. Let's talk about what it involves. There are many occasions when we want to predict things. What is the chance of rain tomorrow? How many people will buy one of your products? Is this email spam or not?

Machine learning is the art of making predictions using "models" that take input data, run calculations, and then output answers. There are several types of predictions that machine learning models can make. Broadly speaking, they can be summarized into two types of machine learning algorithms:

• Supervised Learning:  Where algorithms learn patterns from existing data and apply them on new data (e.g., forecasting future stock prices based on historical ones)
• Unsupervised Learning: Where algorithms discover general patterns in the data (e.g., analyzing customer data and segmenting them based on behavior)

Below, you will find typical examples of different types of machine learning algorithms and some of the questions they can help you answer.

For each type of prediction, several types of models can be used to make that prediction. You can check them out by taking a look at this cheat sheet.

Another type of model is called a neural network. It’s inspired by the brain's structure, consisting of neurons connected in a network. This type of model is notable for its use in deep learning.

Deep Learning: Tackling Complex Cognition Tasks

As explained earlier, neural networks are inspired by the brain's structure. The neurons in a neural network are organized into "layers." Each additional layer in a neural network allows for more calculations to output the best possible results. However, that does not come without its complications. 

The number of layers in a neural network is called the "depth" of the network. A neural network with lots of layers is called a deep neural network, and deep learning is a subset of machine learning where the model uses a deep neural network.

Deep learning is important because of its ability to deal with more complicated “cognition” tasks such as image recognition, self-driving cars, and more. Deep learning models perform better than machine learning models on unstructured data types like text or images. 

They also need less help with "feature engineering"—a technique for processing the data to get it ready to be input into the model. The artificial intelligence examples shown before, such as chatbots, virtual assistants, and image recognition, are all powered by deep learning models.

Data Science: The Art of Science of Making Data Useful

Like artificial intelligence, data science is a discipline rather than a set of tools. It is different from artificial intelligence in that it involves a variety of tools and techniques, including exploratory analysis, data analysis, data visualization, and more, to make use of data. Moreover, data scientists work on much more than just developing machine learning models. 

For example, a data scientist tasked with understanding fraudulent transactions at their company may use a machine learning model to make predictions about fraud. However, they would also use statistics to try and understand the drivers behind what causes fraud, run experiments to try and reduce fraud, and then provide a report to management with visualizations that help the audience make decisions.

The cross-disciplinary nature of data science makes it one of the most attractive career paths out there, especially given the imbalance between the supply and demand of folks with this skillset. Check out the following infographic to learn more about the data science career path. 

Want to Learn More?

We hope you enjoyed this short introduction to artificial intelligence, machine learning, deep learning, and data science and that you feel more confident in understanding their commonalities and differences. 

In the next entry of the series, we’ll be looking to uncover the four types of analytics when working with data and outline some of the most fundamental statistical notions in data today. If you want to start your data learning journey today, check out the following resources. 

• Get started with our Understanding Data Topics Skill Track
• Subscribe to the DataFramed Podcast
• Check out our upcoming events
• Read our next entry of the Data Demystified Series on the four types of analytics
• Read our guide on how to become a machine learning engineer
• Learn AI with our full catalog of resources