Motivation & feedback from the previous course: Together with Johannes Hötter, our first online course on openHPI around the topics of “Artificial Intelligence and Machine Learning for beginners” reached over 15,800 learners (as of June 2022). In a previous blog post, I reflected on the experience of conceptualizing, recording, and hosting our first online course.

At the end of the previous course, we tried to understand how the participants perceived the course, how we could improve, which questions were still open, and if participants wished for a continuation. For that, we used the “I wish, I like” feedback method in a post-course survey, which is well known, for example, in the field of design thinking. The main feedback points from the first course are the following:

  • Good conceptual overview in the introductory course: A lot of feedback focused on the good general overview and that we conveyed important concepts and terminology well. However, some participants remarked that this depth and the covered topics were already sufficient.
  • More practical examples with real data & code: A key feedback by participants was that the intro course could benefit from more practical examples. This is very valid feedback, as we partly used very simple toy examples or metaphors to make the concepts easy to understand. It was also desired to have code to try out for yourself.
  • Higher level of detail and more content: Although our first course was already designed for four weeks, some points remained unanswered or not dealt with - that means also some feedback aimed at a higher level of detail and additional content.

Overall, we this was great feedback. But since there is still so much ground to cover and so many fascinating topics - Johannes and I decided to also create a second course. This time focussed on practical examples.

The organization: The course is organized in four weeks, containing both video and textual lectures and self-quizzes after each video. After each week, there is a graded homework quiz, and after week 4, there follows the final exam in quiz form.

Unlike the first course, we decided to focus on individual projects and provide content per week alongside a concrete use case. Each week starts with a look into the dataset and dive deeper into some theoretical topics which are translated to the specific dataset & application. Then the results are analyzed and interpreted. In this course, we also showed practical code examples, for which we used Jupyter Notebooks and the RISE extension to be able to present the notebooks as presentations. We have published the notebooks on Github in a repository and used Google Colab so that people without extensive experience in coding can also try out our code without installation effort.

When preparing the content for this second online course, Johannes and I were left with so many topics not covered that fascinate us. We, therefore, chose to have two “hot topics” per week as an excursion. The “hot topics” covered are Reinforcement Learning with OpenAI Gym, ML Model Governance, AutoML, Federated Learning, Large Scale Language Modelling, Data Labelling, Bias and Fairness in ML and Explainability (XAI). The weeks are organized as follows:

  • Week 1: Reiteration of important concepts and project on housing price prediction.
  • Week 2: Practical project on film recommendations.
  • Week 3: Sentiment analysis in movie reviews.
  • Week 4: Recognition of sign language images and translation into text.

The official course period was October 6th 2021 - November 3rd 2021, but the online course is available for free self-paced learning anytime on openHPI.

The target audience: In essence, we tried to target people with little to no mathematics and programming background. As we did not include exercises in which the participants need to code, we did not presume any programming skills. Interestingly, 78% of roughly 2000 participants who attended our online course and filled out the pre-course survey already had programming experience.

Learnings: With our second free online course on openHPI, we reached over 7,000 learners and an average of 4.75/5 stars with 48 ratings (as of June 2022). Compared to the first course, fewer participants registered. We expected this trend as we narrowed the target group to people who wanted to go beyond the first course.

  • Practical exercises and code. Even if the point may seem obvious - but as we included a practical part and code for the second online course - it means a significant additional effort that we underestimated to some extent. Not only in preparation but also in the number of questions and problems that did come up.
  • Learning by teaching. I guess the statement “you best learn by teaching others” is very well known. After both courses, this is something that I really would stress. One must really understand the material 100% and also “withstand” tough questions and not just know a topic superficially. This becomes especially clear when you realize that several thousand people see what is said and that this is recorded on videos. So for us, it was also a great learning experience in itself - as a refresher or now understanding some topics in more detail.
  • Teaching is fun. I really liked the preparation, hosting the course, and the interaction with learners in our forum. This has encouraged me to continue to share knowledge - even if I have not decided, if that will be explicitly in the form of an online course or on-the-job such as occasional knowledge-sharing sessions. Trying to look through the eyes of the audience, conveying knowledge and last but not least, share your fascination with the topic. This also describes what I remember of my favorite university professors or teachers back in my school life. And I find that a great purpose and a great thing to strive for.

Final remarks: If I reflect on the second course, I would summarize that we just continued as it was so much fun the first time. We had so many points that we still wanted to convey. It was a nice challenge and a lot of fun. Designing the online course from the first minutes and how we structure the weeks, what content we cover, to preparing the slides and the programming examples to shooting the videos and supervising the course and the forum during the main course runtime. I guess someone who attended both courses also wrote that as feedback - “it was visible how we became more professional along the way”, which is great feedback.

We were thrilled that we were able to expand our reach with the course further, and we were for example invited to two podcasts because of the two courses:

Summary and outlook: Even though our second course scores below our first if we look at the hard facts and numbers, we are still very proud of what we have achieved as a hobby project alongside our studies and working/founding and leading a startup. Overall, we want to “keep the momentum” even if we still have to decide in which way or form - stay tuned :)