Andrew Ng: the professor who put machine learning online for everyone

A researcher who loved to teach

Andrew Ng studied at Carnegie Mellon University for his bachelor's degree, at MIT for his master's, and at the University of California, Berkeley for his PhD. He then joined Stanford University, where he later became Director of the Stanford AI Lab, a home to more than twenty faculty members and their research groups. He has written or co-written more than 200 research papers in AI and related fields.

At Stanford, his Machine Learning class usually had around 350 students. From 2008, its video lectures were put on YouTube, and more than 200,000 people watched them. That was a hint that many people outside Stanford wanted to learn the subject too.

“the far more important purpose of quizzes and homeworks is not to test the student but to give students the opportunity to practice with the material and thus remember the material longer.”

— Andrew Ng, via NPR Talk of the Nation (via WWNO)

The class that went to the world

On 15 August 2011, Ng uploaded an introduction video for a free online version of his Machine Learning course. The New York Times wrote about it, along with two other Stanford courses, the same day. This time the course was interactive, with short videos, quizzes, programming exercises and a grade.

In a 2012 interview with NPR, Ng said that within two days thousands of students had signed on, more than he had taught in his whole life, and that over the following months the number grew to about 100,000. He worked out that to reach that many people in a normal classroom, he would have had to teach his Stanford class for 250 years. A technology news site reported that 104,000 students enrolled in that autumn 2011 run and about 13,000 completed it.

The course was built for busy learners. Videos were cut into pieces of around ten minutes, with questions placed inside them. Students wrote real programs in a language called Octave, and the programming work counted for two-thirds of the score.

Coursera, Google Brain and beyond

In 2012, Ng and his Stanford colleague Daphne Koller turned the experiment into a company, Coursera. Ng told NPR that it was very important to both of them that the courses be free, because many of the people who needed them most, especially outside the United States, might not have even a dollar or a credit card. The machine learning course was reworked for Coursera in April 2012.

Ng was also the founding lead of the Google Brain team, which helped Google become a modern AI company. In May 2014 he left his full-time role at Coursera to join Baidu as Vice President and Chief Scientist, where he led a team of about 1,300 people working on AI.

In 2017 he founded DeepLearning.AI to teach AI more widely. On 15 August 2017, exactly six years after his first online class, the first courses of his Deep Learning Specialization went live, this time using Python and notebooks that run in the browser. His official website says more than 8 million people have taken an AI class from him, and in 2023 he was named to the Time100 AI list. He remains chairman and co-founder of Coursera and an adjunct professor at Stanford.

The skills and habits that made the difference

The first was breaking a hard subject into small steps. Ng's courses are known for short videos, simple examples and one idea at a time. Machine learning can look like a wall of mathematics, but his teaching treats it as a staircase.

The second was believing in practice. In the NPR interview he explained that quizzes and homework matter less for testing students than for giving them a chance to practise and remember the material longer. His courses made learners write code, not just watch.

The third was keeping the door open. From YouTube lectures to free online classes to Coursera, he kept looking for ways to let more people learn. He also kept learning himself, moving from classic machine learning to deep learning and building new courses as the field changed.