I often get requests like: “I would like to move into AI knowledge and applications. Can you please advise on what is the optimum learning path that I could leverage optimally?”
I am happy to do this! So here is my – obviously very subjective – advice:
Get your hardware and software gear ready
You can start with your notebook or PC. Install Jupyter, it is easy and intuitive to play around with and a lot of fun! Later, start to look into PyCharm for bigger projects. For most of the tutorials and smaller projects, your laptop will be totally fine. Later, if you train bigger models and bigger data sets, you could move to a GPU on an (NVIDIA) graphics card from your gaming PC or you use virtual machines from one of the big cloud service providers. Using them on an hourly basis is not really expensive.
Learn Python
Machine Learning (ML) is a practical engineering-oriented field. You cannot learn it from books and need to get practical experience. Python is the best language to start with: The most used language in ML and fairly easy to learn.
Start learning ML with tutorials or online courses
There are numerous tutorials and online training classes available on the internet. Follow one of them and implement simple examples and use cases with ML frameworks like Keras, TensorFlow, PyTorch or scikit-learn. It is really important that you get some practical experience.
Start your own project
Choose an example you are really interested in and care about. It can be an example from your company or from your private environment. Think of a simple chatbot, object recognition with your laptop or phone camera, sentiment analysis of documents or texts, or working on an easy kaggle.com competition. Choose anything you are passionate about enough to finish it!
Get a deeper understanding
If you have a mathematical background, you are lucky! Mathematics is the backbone of ML and the secret of its success! So if you are adventurous, dive into the mathematical foundation of ML like statistics, probability, linear algebra, and gradient descent. A great book to do this is Deep Learning from Ian Goodfellow, Yoshua Bengio, and Aaron Courville, available online at https://www.deeplearningbook.org/. The understanding of the mathematical background helps you to optimize models or invent new ones. If this is not your world, don’t worry: ML is pretty much an engineering field and you can achieve quite a lot with good practical experience. Other great sources to deepen your knowledge are classes of top universities, like the CS* classes from Stanford University. The class recordings, the assignments, and all other class materials are available online free of charge.
Exchange and learn with others
Don’t do this alone. It is much more fun and motivational to do this in a group of like-minded people. Look around: Maybe there are colleagues you could start a learning group with. In many cities, there are regular AI MeetUps or events like School of AI or Hackathons. And if not, start your own one!
Code every day
You do not learn ML only out of books. So it is a great habit to code every day. Nobody can put this better than Geoffrey Hinton, Turing prize winner, and one of the “godfathers” of ML: “Reading rots the mind. Practitioners should figure out how they would solve a given problem, and only then read about how others solved it.”
Most important: Don’t give up and have a lot of fun on your journey into the fascinating world of ML and AI!
Christoph Windheuser, Global Head of AI at ThoughtWorks Inc.
(this article was originally published on LinkedIn)