Abstract:

The massive open online course Elements of AI developed by the Helsinki University and the technology company Reaktor has been an international success story in educating first the Finnish and then the world population about the basics of AI. Sweden was the second country to adopt the course on a national scale in May 2019. Linköping University has been the academic partner and has offered university credits for the course. Since the launch more than 32000 people have started the course, more than 6000 have finished it and more than 2600 have taken out university credits for the course. World wide, more than 640 000 people have started the course and more than 75 000 have finished it.

This paper describes the course, how to set up and run a university course with thousands of students running across the semesters with continuous admittance and examination, and the lessons learned.

Elements of AI gives a broad overview of AI to a general audience. Besides machine learning, it also covers search-based problem solving and the history of the different paradigms in AI. The course goes one step beyond the buzzwords and provides concrete examples of AI techniques such as neural networks, minimax, and Bayesian techniques, as well as the limitations of different methods. One of the six chapters is dedicated to the societal impact and ethics of AI.

From a university perspective, the course is examined in two steps. The first is to complete the online material and the second is to upload the certificate and do a validation test on the university learning platform. The validation test consists of six questions of the same character as in the online material. A new test is provided every month.

The course has a very diverse set of students. The oldest is over 80 and the youngest has just finished high school (you need a high school degree to be admitted to the university course). We have had whole families taking the course and companies educating a significant part of its staff. More than 100 companies have for example pledged to take the course. This has been extra important during the pandemic when many people have had extra time for professional development.

To manage a course with a few thousand students every semester most of the administration has been automated. Initially one faculty (the first author of this paper) worked about 10% with the course and one PhD student level course assistant worked about 60%. When the course was up and running, most of the day-to-day work was managed by three Master's level students working 20% each. Currently it is managed by one Master's level student working 20% with the course assistant supporting him (mainly by maintaining and improving the scripts that powers the reporting) and the faculty responsible for the examination and final reporting.

The university course gets very good evaluations. For 2020, the average score was 4.4 (out of 5), which is very high. The course got 4.65 on the question whether the content provided what is needed to reach the course goals and 4.55 on the question of whether the examination was relevant for the course. It is clear that the course is highly appreciated.

The overall experience of setting up and running a really large scale course has been very interesting and rewarding. The impact has been very high, with many companies asking their employees to take the course. We expect to see similar educational collaborations to proliferate in the future.

Keywords:

AI, inclusion, higher education, life-long learning, online education.