IBM’s “Introduction to Artificial Intelligence,” offered through Coursera, is a short course designed for true beginners. As the architects of Watson, the AI of Jeopardy! fame, IBM is a market leader in AI applications and this introductory course can act as a stepping stone for students who wish to gain an entry-level understanding about the subject.
The curriculum is divided into four modules: applications and examples of AI; AI concepts, terminology, and application areas; AI issues, concerns, and ethical considerations; and the future with AI in action. Taught by Rav Ahuja, IBM’s global director for AI and data science, the course can be taken on its own or as part of a longer track to specialization in a more niche area of AI.
Microsoft’s “Introduction to Artificial Intelligence” short course is the first of ten courses in its professional program on AI and it can be taken as a standalone. Students need nothing more than high school math skills and a basic knowledge of programming to participate.
In a mix of hands-on learning and pre-recorded videos, the course covers how software can be taught to process and extract meaning from natural language and images. The curriculum is divided into five modules: machine learning (the foundation of AI); text and speech (understanding language); computer vision (seeing the world through AI); bots (conversation as a platform); and the next steps. Upon graduation, students will have learned how to build simple machine learning models, how to use Python and Microsoft cognitive services, and how to implement conversational bots through the Microsoft Bot Framework.
DeepLearning.AI is a new educational venture by Andrew Ng, the co-founder of Coursera and a thought leader in AI. Their short course “AI for Everyone” is designed to give an overarching and fundamental understanding of what AI is and what its applications are. There are no prerequisites.
The program is divided into four modules: defining AI; building AI projects; building AI in a company; and AI and society. Students learn what AI is and what it’s capable (and incapable) of doing. Furthermore, they’ll learn how to spot opportunities to apply AI in their own organizations and how to deploy them while considering ethical issues, project logistics, and corporate strategy. While largely non-technical in nature, engineers can and do take the course in order to better understand the business applications of AI.
Harvard University (edX)
Harvard University’s short course entitled “Data Science: Machine Learning” teaches students the science behind recommendation systems. As one of the most popular applications of AI and data science, recommendation systems provide a convenient entry point for those wishing to learn the fundamentals of machine learning.
The course, which is designed for relative beginners, uses recommendation systems as a way to explore the basics of machine learning, including fundamental tasks like how to perform cross-validation to avoid overtraining and how to implement regularization. Taught by Harvard Professor Rafael Irizarry, an expert in applied statistics, the curriculum teaches students how to train data and discover predictive relationships while building their own recommendation model.
Google (Google AI)
Google’s “Machine Learning Crash Course” bills itself as a fast-paced and practical introduction to the world of machine learning. The fully-online curriculum is a blend of real-world case studies, tasked exercises, interactive visualizations, and lectures from Google researchers.
The course consists of 25 lessons, each of varying lengths, and covers topics that include: framing; reducing loss; first steps with TensorFlow; generalization; logistic regression; classification; and training neural networks. At their own pace, students learn how to determine whether a model is effective, how to represent data so that a program can learn from it and how the finer points of machine learning work. While the program does classify itself as an introductory crash course, students are expected to have a mastery of intro-level algebra, proficiency in programming basics, and some experience coding in Python.
Stanford University (Coursera)
Stanford University’s online course in machine learning is one of the most highly rated on the Coursera platform. Designed and taught by Andrew Ng, adjunct professor of computer science at Stanford University, it’s this course that led to the creation of Coursera itself, of which Ng is also a co-founder. Ng’s work on machine learning is legendary, leading to his development of Google Brain, a project that builds massive-scale deep learning algorithms.
The course is divided into seven modules: linear regression with one variable; linear algebra review; linear regression with multiple variables; Octave/MATLAB tutorial; logistic regression; regularization; and neural networks and representation. The curriculum covers a lot of ground, from a broad introduction all the way to best practices in Silicon Valley relating to machine learning and AI.
University of Washington (Coursera)
The University of Washington’s short course in machine learning fundamentals takes a case-study approach to the subject. The program is taught by Carlos Guestrin and Emily Fox, professors in machine learning at the University of Washington. Guestrin specializes in computer science, while Fox specializes in statistics and they both hold multiple national awards in their fields.
Delivered through the Coursera platform, the curriculum is divided into the following topics, which are each accompanied by their own case study: regression (predicting house prices); classification (analyzing sentiment); clustering and similarity (retrieving documents); recommending products (Amazon and Netflix); and deep learning (searching for images). As part of a larger four-course specialization in machine learning, this introductory short course can be taken as a standalone or used as a stepping stone into the deeper aspects of the subject.
Nvidia, a market leader in the manufacturing of graphics processing units (GPUs), offers a short course in the fundamentals of deep learning for computer vision through its Deep Learning Institute.
Computer vision is a subdiscipline of AI that allows computers to process visual information in a manner similar to humans and Nvidia is uniquely positioned to lecture on the subject due to its GPU market dominance. The curriculum covers topics such as deep neural networks; GPUs; big data; categories of performance; deployed pretrained networks; and beyond image classification. Throughout the course, students learn how to implement deep learning workflows; to train a deep neural network to correctly classify images it’s never seen before; and to experiment with data to optimize performance. By the end, students will have constructed and deployed their own neural networks capable of solving real-world problems.
IBM, in association with Coursera, offers an advanced short course in deep learning with Python and PyTorch. It’s taught by Dr. Joseph Santarcangelo, an IBM data scientist whose research focuses on machine learning, signal processing, and computer vision. Prerequisites to this course include familiarity with Python and Jupyter notebooks, machine learning concepts, and deep learning concepts.
The curriculum is divided into the following modules: an introduction to PyTorch; linear regressions; classification; neural networks; deep networks; and computer vision networks. Overall, students learn how to develop deep learning modules and gain a working understanding of deep neural networks, convolutional neural networks, and transfer learning.
Microsoft has a short course, entitled “Deep Learning Explained,” that teaches students an intuitive approach to building complex models for machines to solve real-world problems. This is an intermediate course that requires students to have basic programming skills, a working knowledge of data science, and familiarity in using Python specifically for data science applications.
Taught by Dr. Steve Elston, who holds a PhD in geophysics from Princeton and has led several multinational data science teams across the private sector, the course is designed to give students a fundamental understanding of deep learning concepts and then teach them powerful motifs that can be used flexibly in building end-to-end neural networks.
The curriculum is divided into six objectives: learning the basics of deep learning and recapping machine learning concepts; building a multi-class classification model using logistic regression; detecting digits in a handwritten digit image through end-to-end models and deep neural networks; improving the handwritten digit recognition with a convolutional network; building a model to forecast time data using a recurrent network; and building a text data application using recurrent LSTM units.