Crash course: local AI in education: control over your own data

Last week I, Pascal Mariany, had the pleasure of giving a workshop at the well-attended Samen Digitaal Café on AI at Hogeschool Utrecht. With more than 80 registrations, the interest in local AI solutions was terrific! In two sessions, over 60 participants received an intensive crash course on the benefits of local AI and how this technology can be used safely and effectively in education. As a digital goody bag, the presentation can be downloaded in this blog post!

The rise of local AI: why it matters for education

The workshop began with a short introduction to the new European AI Act. This legislation places strict requirements on the transparency, safety and privacy of AI systems used in education. Cloud-based AI solutions bring risks, especially in the area of data privacy and control. Local AI offers a way out here, with educational institutions keeping control over their own data, which stays safely within the walls of the school or university.

Crash course local AI: what do you need?

During the workshop I discussed the essential hardware and software for running AI models such as LLaMA locally yourself. For those who want to get started, there are some basic requirements: a GPU with at least 8 GB of VRAM, 32 GB of RAM, and preferably an SSD for fast data storage. But MacBooks from an M1 chip with 16 GB of RAM also run local LLMs perfectly well! With this setup you can run AI without depending on external servers, which not only increases security but also contributes to a better learning experience for students.

Examples from practice: from feedback to peer reviews and much more!

Local AI can be used in education in countless ways. Think, for example, of the Feedback Companion, which can automatically generate feedback based on learning goals. AI can also contribute to peer reviews by providing structure and encouraging constructive feedback. During sprint reviews or retrospectives, local AI provides insight into progress and suggestions for improvement. These are just a few of the many applications discussed in the session, because you can also, for instance, analyse X-rays or have local AI models program a Snake game.

Try it yourself: a knowledge clip series for more depth

For those who want to know more, I am happy to refer you to the knowledge clips on OnderwijsOrakel.nl. In them I explain step by step how to set up, manage and effectively use local AI within your educational context. Watch the knowledge clips to get off to a flying start: Knowledge clips – Local AI.