In the mid-2000s, AI was still just a curiosity to the world at large. At Stanford University, however, one of the most popular classes on campus was Andrew Ng’s CS229 machine learning course. Enrollment was frequently too large to fit in the classroom, yet he wanted even more people to be able to master machine learning.
So, working with a few students, he created an online Machine Learning course that could be taken by anyone with an internet connection and a desire to learn. The rest is history. Coursera launched in 2012 with Machine Learning as its flagship title. It was also the platform’s most popular, with almost 5 million enrollments.
This year, to celebrate the course’s 10 year anniversary, DeepLearning.AI and Stanford Online released a successor — the Machine Learning Specialization. Andrew spoke with us about how the new Specialization improves on the original, who should take it, and how it fits into the modern AI builder’s career arc.
Why did you decide to create a new version of the Machine Learning Specialization?
Andrew Ng: The field of machine learning has advanced considerably since the course debuted 10 years ago. While the foundations of machine learning change slowly — which is why the original course has had a long shelf life — I was also eager to update it to make sure it reflects the most important concepts in today’s machine learning as well as the latest software tools.
How does the new Machine Learning Specialization differ from the original machine learning course from Stanford?
The new Specialization has been significantly updated:
- • It includes an expanded list of topics that focus on the most important machine learning concepts (such as modern neural networks and decision trees) and tools (such as TensorFlow).
- • We redesigned the assignments and lectures to use Python — the programming language of choice for machine learning developers — rather than Octave.
- • Unlike the original course, which required students to have a more advanced background in math, we designed the new Specialization to be more accessible for first-time students of machine learning.
- • Also new are ungraded code notebooks with sample code and interactive graphs. These will help you visualize what an algorithm is doing and make it easier to complete programming exercises.
- • Using the emerging best practices from the past decade, we have updated the section on practical advice for applying machine learning to real world projects.
Should people who took the original machine learning course consider taking the new Machine Learning Specialization?
If you completed the original Machine Learning course and still feel like the topics covered are fresh in your mind, then probably not. Instead, I recommend taking the Deep Learning Specialization, which is the most common next step for people who want to continue developing their skills after completing Machine Learning.
If, however, the material from the original Machine Learning course isn’t fresh in your mind, this new Specialization will help you strengthen foundations that you will return to throughout your career. You’ll be able to quickly breeze through the parts you are already familiar with, and the attention you give to parts you are less familiar with will be time well spent!
What does the new Machine Learning Specialization offer to someone who is new to AI or machine learning?
This Specialization is beginner-friendly, and is accessible to someone who knows only very basic programming and high school math. The Machine Learning Specialization will set you well on the path to being able to build and apply your own learning systems.
When I look back on my own career, I think the best investment I made was to spend time developing a really good understanding of the foundations of machine learning. The Machine Learning Specialization will help you to do the same.
What skills should people have before they take the Machine Learning Specialization?
You should understand some basic coding — for example, you should be familiar with function calls, variables, for loops, and if statements. You should also know basic high-school level math, including arithmetic and algebra. For example, if 5x+1 = 11, can you solve for x? Any more advanced math will be explained along the way. In particular, don’t worry about knowing any calculus, linear algebra or statistics.
What really matters is a desire to learn about AI and ML — we’ll work with you on the rest!
How can I transition into a machine learning job after completing the Specialization?
After completing the Machine Learning Specialization, start working on more projects and building up a portfolio of work. Putting the completed projects on your resume will make you more attractive to hiring managers, and the things you learn will make you more confident during interviews.
You should also continue taking courses. The most common second course to take is the Deep Learning Specialization, which will give you a strong foundation in neural networks, which are a huge fraction of machine learning today. Further coursework will also help you narrow down a subfield within machine learning that you might want to work in. For example, if you are looking for jobs in healthcare, look into AI for Medicine. If you want to join in with all the exciting stuff happening with large language models, check out Natural Language Processing.
Once you are ready to start applying for jobs, don’t be shy about reaching out to people you met while building your portfolio and tell them what kind of position you’re looking for. Another smart tactic is to look on LinkedIn for professionals who are working at companies you’d like to join. Many will be happy to answer your questions, critique your portfolio, and offer advice on how to present yourself as somebody worth hiring.
This sequence of letters (part 1, part 2, part 3, part 4) in The Batch also dives deeper into specific career advice.
How can I find projects after completing the Machine Learning Specialization?
Many people’s first project involves applying machine learning to a hobby or a fun side-project (like predicting if your favorite sports team will win, or using a machine learning based script to help a colleague automate some of their work). A career involves a sequence of projects, hopefully growing in scope, complexity and impact. So when you are starting out, it is completely fine to start small, and use that to learn and continue to grow!
One additional way to find meaningful projects is by looking for ways to apply these ideas at your current job. Even if you don’t currently work in machine learning you may be able to find some data to analyze or process to automate.
If you are having trouble finding inspiration, you can also look on discussion boards, such as Discourse, for like-minded people who are seeking help with their own various machine learning projects (click here to sign up for Discourse; existing users can sign in here). Introduce yourself and take on tasks that fit your current skill set and also allow you to slightly stretch your skills. You’ll gain practical knowledge, and the people you work with will become part of your professional network.
And remember that it is completely fine to start small! That’s how you’ll gain skills that will build to more significant endeavors. Achieving a quick win also helps you keep your motivation up and builds forward momentum. Learn by doing, and then share what you’ve learned on Github, LinkedIn, or Discourse.
What was your motivation for creating the original Machine Learning course ten years ago?
For years I had been teaching my on-campus machine learning course at Stanford University, reaching about 400 students a year. I noticed that I was walking into the same room and giving the same lecture year after year — even telling the same jokes — and decided this wasn’t the best way to serve students. So I started exploring alternatives.
One of my first projects was Stanford Engineering Everywhere (SEE), which posted videos and materials from my machine learning course and a few other Stanford courses on the internet. I received a lot of very positive feedback, which encouraged me to keep going.
I worked with a few Stanford colleagues and together we iterated through many different versions and features. In 2011 we finally hit on a recipe for scalable online courses. Details on this story are here.
One rule that I’ve tried to stay true to throughout this process is to always put learners first. My teams also use this as a guide whether they are designing curriculum, and or deciding whether to implement some new feature, like the 1.5x playback tool.
Since 2011, I’ve been thrilled and honored at how the online education movement has been embraced: By numerous students, a global community of educators, the teams at Coursera and DeepLearning.AI, and their partner institutions. Even though we’ve come a long way, I feel we’re still in the early days of online education, and specifically that there’s still a lot of technical and pedagogical innovation ahead!
This is great. How do I get started?
There’s no time like the present. I hope you’ll jump in and enroll in the Machine Learning Specialization!
Ready to #BreakIntoAI with the Machine Learning Specialization?