A teacher asked if we had a mute-all button. We did not
The hardest part of putting AI into schools is not the technology. It is that you do not know what a teacher standing in front of a class is actually worried about.
After one product discussion, a homeroom teacher asked me, almost in passing: "Fifty children in a classroom is already loud enough. If fifty AI companions all start talking too, can I still keep control of the lesson? Do you have a mute-all?"
I could not answer, because the feature was not on our list.
The real barrier is more mundane than you expect
People assume the biggest obstacle to AI in schools is technical. It is more ordinary than that. Many schools do not allow electronics with screens in the classroom, because they pull attention away. That single rule eliminates most devices before any capability discussion starts.
Our way around it was to build screenless, voice-first hardware. Students interact by talking, there is no display, and teachers can accept it. That is not a technical breakthrough. It is having thought properly about where the thing gets used.
The knowledge base is the second gate
A teacher will not accept an AI that can say anything sitting in their classroom. If a child asks something odd and the AI improvises, who carries that?
So the device runs a knowledge base and its answers stay inside a defined boundary. The teacher knows what it can and cannot talk about. They need that boundary to exist before anything else matters.
The knowledge base itself is not exotic technology. The hard part is that you have to define what belongs in it with the teachers, and that process is harder than writing the code.
Mute-all is not a mute button
After the teacher raised it, I spent a long time going back over our planning. We had been preoccupied with answer accuracy, latency, and how fast knowledge updates. Nobody had thought about fifty devices speaking at once in one room and what the teacher does about it.
Only somebody who actually stands at the front of a classroom asks that question.
Turning it into a feature was not a matter of adding a mute key either. It needed decisions on several things.
- Does the teacher control it from a console, or is there a physical key on each device
- Once muted, does the device keep listening, which trades battery against instant response
- Can one device be muted individually rather than all of them
Every one of those depends on how a lesson actually runs, and we went back to that teacher repeatedly before settling the design.
What is actually scarce
Two years into building AI hardware I am increasingly sure of one thing. Models can be bought and compute can be rented. The few dozen real scenarios held by a practitioner in the field are the part you have to go and dig up yourself.
Mute-all came from there. Before it, how to draw the knowledge boundary was also taught to us by teachers, hands on. After it there will be more requirements we would not have thought of.
If you build education, training or children's products, that is the collaboration shape worth having. You bring the real usage problem, and we work out how to solve it technically.
Related
- 4G AI Learning Companion — one-press class start, batch mute, seating chart
- Request a sample for your school
Want to try the hardware behind these notes?
Six AI hardware platforms are in mass production. Tell us your market and channel and we send an evaluation sample.
Request a sample