Repetition Isn’t Learning
Our text today is a passage from the introduction to Difference and Repetition by Gilles Deleuze:
We learn nothing from those who say: “Do as I do”. Our only teachers are those who tell us to “do with me”, and are able to emit signs to be developed in heterogeneity rather than propose gestures for us to reproduce.
— (Deleuze (2014 [1968]), Difference and Repetition, Bloomsbury Academic edition, p27).
One of the ongoing challenges in education (in 2025 as much as ever) is the belief that repeating information is the same as learning. Some students think that restating what they’ve been told — or what AI tools can regurgitate on their behalf — is enough. And to be fair, plenty of educators still treat learning as something that happens by talking at people.
But education isn’t about perfect repetition. That’s what AI already does: generating the most statistically likely stream of words on a topic. That’s not learning; it’s mimicry.
Real learning isn’t static — it’s active, messy, and context-specific. It’s not about knowing what was said, but understanding why it matters and how it fits into your world. It’s not just knowledge-as-being (something fixed); it’s knowledge-as-becoming (something alive and in motion).
The problem? Much of our education system still mimics a production line — predictable, standardised, and focused on measurable outputs. But learners aren’t products. They engage differently, bring their own contexts, and make meaning in unique ways.
That’s why I’m drawn to methods like Training from the Back of the Room (TBR), which treat experience as central to learning. It’s about doing something with the knowledge — testing it, reshaping it, applying it. One technique TBR promotes is the “teachback”, where students research a topic and explain it to their peers, linking it to their own work. It’s simple, but powerful: they take signs (textbook ideas) and transform them, together, in dialogue, and then teach them to their co-learners.
This is what Deleuze meant by “do with me”. The knowledge isn’t handed down; it’s worked on, questioned, adapted. It becomes theirs, and that variety of difference of understanding that emerges is as distinct as their experiences (“developed in heterogeneity”).
So when I read Jose Casal’s piece on NotebookLM potentially revolutionising learning, I was intrigued. It’s great to see tools evolving. But I kept coming back to the same question: is what we’re seeing really learning?
If a tool just reshuffles existing information into a podcast or summary, it might help someone revise, but that’s not the same as learning. And to be honest, I found the podcast hosts in the demo a bit cringe. Personally, I learn by taking notes, making connections, and actively questioning what I’m reading — something current generative AI still struggles to support meaningfully.
That said, I do use ChatGPT in my own academic work — as a sparring partner, of sorts. It helps me think things through, especially when I catch it hallucinating and push back. That kind of back-and-forth can be useful. But even then, the tool can’t move beyond the information it’s trained on. It doesn’t know if your ideas are good. It can’t assess your learning, or help you make sense of how a concept transforms your understanding.
So yes, I think NotebookLM and tools like it will change education. But if they’re just rearranging content, they’re still rooted in knowledge-as-being. The real transformation will come when they support knowledge-as-becoming — when they help learners make sense of ideas in their own messy, complex worlds.
And this is where an educator — a teacher, trainer, lecturer, supervisor — is still essential, in my view. It turns out that magic statistic word machines can handle knowledge-as-being moderately well, but the point of education is becoming.
