When Andrew Ng announced Deeplearning.ai back in June, it was hard to know exactly what the AI frontiersman was up to. In his time since departing as Baidu’s chief scientist, Ng has been developing a sequel of sorts to his popular introductory machine learning Coursera class. This go around, he is focusing specifically on deep learning.
Coursera’s new Deep Learning Specialization Certificate will feature five courses. These classes, a collaboration between Ng and Stanford grad students Kian Katanforoosh and Younes Mourr, will teach neural networks, backpropagation, convolutional nets and recurrent nets in addition to other core aspects of deep learning.
Students who enroll will also be able to participate in applied deep learning projects to address real-world problems in healthcare, music generation and language understanding. Some of the lectures will include interviews with deep learning developers. Three of the five courses are launching today.
“Demand for deep learning vastly outstrips supply,” Ng explained to me in an interview. “It’s concentrated in a few labs, a few leading universities and technically lots of knowledge is in ARXIV.”
Ng’s initial machine learning class became Coursera’s first class back in 2011. Since then, more than 1.5 million students have taken the class. His goal is still helping to create the easiest way for students to contribute to the growth in the field of AI and its applications in addressing global challenges.
“With the rise of electricity, we saw the rise of the electrical engineering discipline,” Ng explained. Ng has long been a proponent of comparing AI to electricity.
A few companies will be providing tools and resources to students, but the aim was to ensure the class didn’t end up over-branded. In recent months, data science communities like Kaggle, which was acquired by Google, have drawn a lot of attention from cloud service providers.
It doesn’t benefit anyone in the long run if students only get exposure to a single deep learning framework from a large tech company because of a sponsorship agreement. But Ng is clear that he’s here to teach the most useful material to learners. In this case, that means teaching deep learning in a low-level language like Python before even thinking about frameworks.
Students interested in taking the new courses should look to the Coursera site here. The only prerequisite is basic programming knowledge, a proficiency in Python and ideally a general understanding of machine learning.
Ng told me that he is working on three distinct projects. While he opted not to specify details, he did say the others would be more specific AI products. And as to funding — Ng thus far didn’t want to get into details.