From Burgdorf to Paris-Saclay: AI & robotics immersion for German high school students

On Friday August 29, 2025, we had the pleasure of welcoming a group of seven students from the BBS Burgdorf vocational high school (near Hanover), accompanied by their teachers, to our NeuroPSI premises and then to the Paris-Saclay campus, Herr Holger Buchmann and Pascal Ströhlein. At the controls on the Learning Robots side : Thomas Deneuxour president and guide for the day.

The day was part of an Erasmus trip to Paris, conceived and supported by Herr Buchmann, an enthusiastic customer of our solution in Germany. The aim was to give his pupils a hands-on experience of artificial intelligence through robotics, to test our latest educational activities and, more broadly, to open up prospects for vocational high schools across the Rhine.

‍

Who are our visitors?

BBS Burgdorf is a multi-disciplinary vocational school. The seven students who joined us are in their final year of high school, and represent a number of different subjects: economics, IT and metalworking. This diversity set the tone from the outset: cross-views, frank questions, and immediate bridges between the uses of AI in management, IT and industry.

The teaching team, led by Herr Holger Buchmann and Pascal Ströhlein, has been looking for hands-on approaches to acculturate students to AI for several years. Their conviction, which we share, is simple: you really understand artificial intelligence by practicing it, by seeing an algorithm learn, make mistakes, progress, and by linking each decision to real-world sensors, data and constraints.

‍

‍

Why this visit matters

This meeting is important for several reasons:

• A Franco-German bridge between active pedagogy and vocational training, in the spirit of Erasmus.
• Feedback from the field is invaluable for Learning Robots: we can observe how our activities speak to students from a variety of backgrounds, and measure the concrete impact on their representations of AI.
• A challenge of skills sovereignty: AI is penetrating all trades, from the supply chain to information systems, from machining to predictive maintenance. Equipping vocational high schools with simple, comprehensible and motivating tools is a strategic investment.

"AI is not black magic: it's a set of techniques that can be made visible, manipulable and discussable in the classroom." - The Learning Robots spirit

‍

At the heart of NeuroPSI: where Learning Robots was born

The day began on our premises, in the heart of the NeuroPSI neuroscience research center. This is where Learning Robots was born. Here, we link neuroscience, AI and robotics to learn by doing. Students have seen that our robots are not magic: a sensor sends information, the algorithm decides, the robot acts. Everything is visible and concrete.

After a brief presentation of our mission to democratize AI through practice, we moved straight on to the workshop.

‍

‍

‍

2-hour AI workshop: run, connect, understand

We proposed our classic robot race, followed by a new "2 pixels" activity. Two intense hours to move from intuition to reasoning, then to experimentation.

1) Robot racing: fun but demanding

First highlight: a track, curves, walls... and robots. The instructions seem simple: go as fast as possible without touching the edges. But behind the apparent simplicity lies a realmachine learning challenge.

The students set up the sensors, choose control strategies, test driving policies and observe the consequences in real time. The key concepts quickly became apparent:

• Training data: what situations must the robot experience to learn effectively?
• Generalization: how to avoid having him "learn by heart" a trajectory instead of understanding the dynamics?
• Feedback: adjust, iterate, compare.

The friendly competition that ensues transforms thinking into energy: each team wants to beat its own time, then the group record. This is where educational robotics really comes into its own, uniting complementary skills around a tangible goal.

2) The "2 pixels" activity: AI laid bare

Second time around, and the big novelty tested with the group: our "2 pixels" activity. The principle: a minimalist neural network, with 2 inputs and 5 outputs, designed to drive the robot to avoid walls.

Step A - Wiring by hand

Before pressing "learn", the students establish the connections themselves: what weights should be given to the two inputs? How can these signals be combined to control the motors (left/right, forward, correction) via the five available outputs? In just a few minutes, you'll understand what "synaptic weight" means, because you can set it and feel it: too much weight on the left, and the robot squeezes the edge; the wrong balance, and it zigzags.

This "manual" moment is a revelation. It showcases theAI that can be explained by practice: a decision is linked to the path of information.

Step B - Learning from training data

Next comes the learning phase: the network is provided with examples (sensor measurements → desired commands) and the algorithm adjusts the weights. Students then compare the human solution with the learned solution. They observe:

• the emergence of a decision boundary legible in the 2D space of the inputs;
• how small weight adjustments transform entire trajectories;
• the result of overlearning (when the robot becomes "too sure" of a rare case and is wrong about the general case);
• how noisy data can throw the system off course, and the importance of data quality.

This stripping-down of a neural network to its bare essentials was enthusiastically received: the students found the workshop "enlightening" for a concrete understanding of what goes on in an AI, far from the black boxes.

What they learned in 120 minutes

• A solid intuition of the perception-decision-action loop.
• The ability to link a weight, an input and an output to observable behavior.
• The distinction between tuning a system and training a model.
• Data engineering reflexes (diversity, noise, sample size).
• A taste for measurement: observe, compare, trace, explain.

‍

‍

‍

‍

‍

From lab to campus: Paris-Saclay in the storm

After the workshop, it was off to the Paris-Saclay campus. From the Lumen library, we took in the beautiful pre-storm light, the dramatic skies making the facades even sharper and the trees almost fluorescent. Then, true to Murphy's Law, just as we were about to take a stroll around the tree-lined campus... we got drenched!

This weather interlude had an unexpected virtue: it brought the group closer together. Sheltered for a few minutes, we discussed career paths, future projects, internships and what a responsible AI culture can change in a workshop, an IT department or a warehouse.

‍

‍

‍

‍

‍

‍

‍

‍

What this means for German vocational schools

The day's experience confirmed several pedagogical points of support that are particularly relevant for establishments like BBS Burgdorf:

1. Learning by doing: robotics makes AI a reality. A student of metallurgy understands the impact of an outlier just as he understands a machining defect: by its effects. This is a powerful analogy.
2. Cross-disciplinary skills: data collection and quality, understanding sensors, simple modeling, performance measurement, critical thinking about results, fundamentals that apply to economics, IT and workshops.
3. Explainable and ethical AI: by making the sensor-algorithm-action chain visible, we make technical choices and their consequences (safety, bias, robustness) debatable. This is the basis for responsible AI.
4. Motivation: the dynamics of the race and the "aha!" effect of activity 2 pixels engage students. We learn best when we take action.
5. Accessibility: you don't need a supercomputer to understand AI. A 2→5 network is enough, if well staged, to lay a lasting foundation.

For teaching teams, it's also a lever forinterdisciplinarity: an AI module can bring together an economics teacher, an IT teacher and an industrial trainer. Each brings his or her own language to the table, and the workshop becomes a crossroads.

‍

‍

‍

‍

AlphAI: simplicity, transparency, action

Throughout the day, AlphAI, our artificial intelligence education software, fulfilled its role of making learning readable, manipulable and documented.

• Easy to use: in just a few clicks, you can go from data acquisition to training, and then on to robot simulation.
• Transparency: curves, weights, outputs... everything that counts is visible and can be explained.
• Action: AI doesn't stay on the screen; it pilots a robot. This is where understanding comes in.

Activity 2 pixels enriches this triptych: it takes the breakdown of internal mechanisms (inputs, weights, outputs) a step further, to catch the eye and connect intuition. When the trajectory changes because a single weight has moved, AI ceases to be an abstraction.

‍

‍

‍

‍

‍

‍

Thanks

Many thanks to Herr Holger Buchmann for his enthusiasm and commitment to bringing AI to life for his students, to Pascal Ströhlein for his always stimulating technical exchanges, and of course to the seven students for their curiosity, energy and team spirit. Our thanks also go to Thomas Deneux, President of Learning Robots, for preparing and hosting the event.

In summary: what they take away, what we retain

What students take with them

• A concrete image of what an AI algorithm is.
• Measuring reflexes andcritical thinking.
• Proof from experience that a simple, well-instrumented model can solve a real problem.
• The desire to dig deeper, whether in IT, industrial organization or economics.

What we remember

• The relevance of Activity 2 pixels in demystifying the "neural network".
• The pedagogical value of robot racing as a common thread for talking about data, models and action.
• The importance of equipping vocational high schools with ready-to-use, flexible and explainable resources.

Opening up the field of possibilities

Back in 1950, Alan Turing asked whether a machine could "think". In 2025, we prefer to ask our students another question: when a machine acts, what do we understand about the way it learns? The strength of educational robotics lies in placing this question in the middle of the room, observable, measurable and debatable by all.

From Burgdorf to Paris-Saclay, Friday August 29 was a fine episode in this adventure: making AI an object of shared understanding, in the service of young people's skills and employability.

If you run or teach at a vocational school in Germany (or elsewhere!) and would like to try out our activities at your school, drop us a line. We'll be delighted to work with you to devise formats tailored to your specific needs and constraints.

‍