In the development of LearningML, one of the key challenges was to create an educational tool for teaching Artificial Intelligence (AI) that did not rely on user accounts or external platforms. This was achieved by implementing the Machine Learning algorithms directly in the user’s browser, using the open source Tensorflow.js library for the ML algorithms and mobilenet to encode images using the technique known as transfer learning.
This approach not only simplified access to the tool, but also demonstrated that it is possible to run complex algorithms locally for educational purposes, where large amounts of data are not required. The experience highlighted the importance of designing practical pedagogical resources that foster computational thinking and bring complex concepts such as AI closer to the school environment.
More importantly, it has become a design principle of LearningML, so that all functionalities to be added (new ML algorithms, recognition of more data types, generative AI, reinforcement learning, …) should be done respecting this design principle.
This will allow maintaining user privacy and complying with the RGPD and the ethical guidelines on the use of artificial intelligence (AI) promoted by the European Union.
Translated with DeepL.com (free version)
