Artificial Intelligence vs Machine Learning vs Deep Learning

The business world is abuzz with talk of Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL). Every day brings exciting news of applications that are expanding the types of tasks computing systems can accomplish. These new developments are incredibly exciting, but they also bring new terms every day. Keeping up with the deluge of terms and concepts can be overwhelming. Let’s address the fundamental question: What’s the difference between Artificial Intelligence, Machine Learning, & Deep Learning?

The most basic answer to this question is that Deep Learning is a form of Machine Learning which in turn is a form of Artificial Intelligence. All Deep Learning is Machine Learning, but not all Machine Learning is Deep Learning. Similarly, all Machine Learning is Artificial Intelligence, but not all Artificial Intelligence is Machine Learning.

Let’s explore the most general term first: Artificial Intelligence

Artificial Intelligence (AI)

An AI system mimics cognitive functions usually associated with human and/or animal cognitive function. AI encompasses a wide swath of applications including optical character recognition (OCR), playing chess, self-driving cars, product recommendations, supply chain management, automated trading strategies, fraud detection, and an ever-increasing array of applications.

Interestingly, the definition of AI doesn’t discuss the implementation strategy of AI systems. While all AI implements some type of algorithm, it’s not until we start talking about the definition of Machine Learning versus other categories of AI that the topic of implementation strategy comes up. Whether the system consists of thousands of if-then statements like a conventional rules engine, mimics the biological structure of the human brain, or uses some yet unthought of strategy, if the system is mimicking cognitive function, then it can be referred to as an AI system.

Machine Learning (ML)

Machine Learning is a category of Artificial Intelligence. ML systems are capable of dynamically updating their behavior based off input data. ML systems contrast with conventional, static programs in their ability to leverage statistics on input data to optimize behavior. Machine Learning has become an increasingly important class of AI solutions over recent years because of the increasing volume of data available to businesses. By providing an automated way to leverage data volume to increase task performance, ML systems are extremely valuable in a data-rich world.

Deep Learning (DL)

Deep Learning is a part of Machine Learning. DL is distinguished from other forms of ML by having multiple layers of processing that can represent a hierarchy of features/concepts. Most forms of DL are models that contain Neural Networks (NN) to model these hierarchies. Neural Networks are very loosely based of the structure of neurons in animal brains. Below, we can see the visual representations of the first two layers of feature detection in an image processing Convolutional Neural Network

Deep Learning is currently the state of the art for many computing tasks including image classification, natural language processing, machine language translation, and recommendation systems. Many of the underlying theoretical principles of Deep Learning have been known for decades but have not been practical to implement. The recent advent of cheaply available, massively parallel computation in the form of Graphical Processing Units (GPU’s) and the coinciding increased availability of large volumes of data are what have recently fueled the rapid advances in the state-of-the-art Deep Learning.

By utilizing the theoretical principles of deep learning with these new technological developments, deep learning systems now represent the state of the art in a wide array of areas including:

  • Recommender Systems
  • Predictive Analytics / Forecasting
  • End to End ROI Analysis on Digital Advertising
  • Speech Recognition
  • Images
    • Object Detection
    • Image Classification
    • Style Transfer
  • Natural Language Processing
    • Machine Language Translation
    • Text Classification
    • Sentiment Analysis
    • Document Summarization

Conclusion

Organizations that we work with have been seeing great results unlocking the value of their data with new Deep Learning systems. Hopefully, you now have a better understanding of the landscape of AI, ML, and Deep Learning. Once you’ve got a framework for thinking about these classes of systems, it can make life a lot simpler!

In our next installment in this series, we’ll explore another commonly discussed ML concept: Unsupervised vs Supervised Learning and where each may be appropriate in your organization.

By |2018-10-19T16:59:30+00:00October 11th, 2018|Categories: Machine Learning/AI|Tags: , , , , |

About the Author:

Hiron is ATownData's resident ML expert. With experience building numerous custom models and ETLs for large enterprises, he prides himself on keeping up to date with the latest in the industry. In his spare time, he enjoys playing jazz guitar and competing in Brazilian Jiu Jitsu.