If you have found the letters AI heavier in the classroom this year, you are not imagining it. This note is a small vocabulary shift that has made lessons easier for the teachers we work with.
The phrase we use
For our own work we say General Natural Intelligence (GNI). We describe our project as being on the attainable path toward General Natural Intelligence, natural not artificial. That single change, from "artificial" to "natural," carries real weight in a room full of students.
We do not call our work AI. We do not call it AGI. The reason is precise, not stylistic: the systems we build and teach are grounded in active inference and generative-model reasoning about a learner in an environment (Class E, following Parr, Pezzulo, and Friston, 2022). That family of ideas describes how a natural agent, a person, an animal, a well-designed system, updates beliefs against experience. It is a program about natural intelligence. Marketing borrowed the word "artificial" for language models. We are not doing that work, so we do not use that word for ours.
What this sounds like with students
You do not have to give a lecture. A one-line framing usually lands:
"The chatbot on your screen is a language model. It predicts the next token. What we are studying is different. It is a way of asking how a natural learner, you, updates what you believe when the world pushes back. We call that General Natural Intelligence, GNI, natural not artificial."
That sentence gives students two clean categories instead of one blurred one. It also tells them, without moralizing, that the tool in their pocket is not the subject of the course.
Why the fence matters here
Our AI-authorship fence for classrooms (Class C, see the configuration in the site cluster) applies to authorship of assignments, not to vocabulary. But the vocabulary supports the fence. If we say "AI" for everything, students hear the fence as arbitrary. If we say "language model" for the chatbot and "GNI" for the active-inference work, the fence lines up with the language, and the rule becomes legible: the language model is a source you must cite and reason about; the GNI work is what you are learning to do.
Field context
For teachers who want the wider picture, Themesis has been tracking how the AGI conversation is shifting under the field, and why grounded framing matters right now: The AGI Landscape Just Changed. Our one-line frame in our own voice: field-level context for why choosing GNI language in the classroom is a defensible, current move, not a private preference.
The honesty posture
We keep saying this in every venue because it is the whole discipline: GNI is a working hypothesis. It has growing, evidence-classed evidence. It is being tested in the open. Do not take the claim on faith. Test the build, inspect the gates, and help us find where it fails. That is the sentence we want students to hear us say out loud at least once a term. It teaches them that a serious research program invites falsification, and it inoculates them against the marketing register that surrounds tools they use every day.
What to try this week
Pick one lesson where the letters AI would ordinarily appear on a slide. Replace them. Say "language model" when you mean a chatbot. Say "General Natural Intelligence" or "active inference" when you mean the human-facing work we teach. Notice what changes in the questions students ask. That small edit is often the whole intervention.