The Radio of Narcissus

In a new world dominated by old ways of learning, AI immersion creates new opportunities for education.

Before people learned from books, they primarily learned things from listening to and watching other people. You don’t need literacy to learn how to drive a plough, fill a bucket, follow an animal through a forest, even layout a farm. We got those from the first generations first through listeners and watchers in the generations that followed. And now, digital media has brought us back there. We are all, once again, a generation of listeners and watchers. Audiobooks and YouTube tutorials have a bigger audience than ever, and podcasts and video essays ostensibly serve those who wish to engage in more topics subtler, more oblique. Of course, as these things emerged people expressed as much concern over them as people do about the birth-pang hallucinations of AI. But all that’s past us now.

If traditional schooling sought to fulfill its aims primarily through the book, with heavy backpacks the uniform of commute, it is headphones – the ones that no longer play rock or hip hop – that are the symbol of the counter movement, the return to the earlier age.

To say that they have contributed to a deschooling of a society would go too far, but their influence has, in some ways, started to eclipse that of traditional schooling. From 2020 to 2024, sales of audiobooks grew 14 percent, while print books grew less than one percent. Podcast consumption, during roughly the same period, grew 11 percent to blogs' three percent.

The sound of the human voice, like the written word, is said to be a hot medium (“hot = high-definition, low audience effort”, McLuhan), in that we aren’t called upon to interpret it all, to piece together different pieces, but merely to listen. However, in the world of growth and learning, where repetition and iteration rule development, one can’t be expected to always follow one audio program from beginning to end and extract everything from it that needs to be extracted. The goal for headphone learners, especially in the initial stages of mastering a subject, is mass input, immersion. To learn is to keep listening. To (self)-assess is to rewind.

The ideal method for this type of learning would be one where you just listened, where you tune in as with a radio. Where you don’t think about what to listen, but rather just immerse yourself in it, lose yourself. There have been audio learning programs that have attempted to achieve this, especially in the world of language learning, but outside of it, too.

Imagine that one listened to a radio station devoted to a single subject. The topics covered a range of diffulty, starting with beginner-grade material, but following with more intermediate and even advanced material. Were one to spend a significant time listening to this station, whether while driving one’s car, cooking dinner, or resting after a difficult day, you’d learn, in a sense, without effort.

Of course, the issue with this kind of radio method is simply one that is common to all forms of mass media. It’s not terribly personalized. And while this may not matter so much in the realm of entertainment (or maybe not as much as people think it does), it does matter in the world of education – the brain is particular about what it needs to learn, in many cases – and random input, especially if it’s too high or too below the level, quickly becomes boring. This 24-hour radio station might be effortless, but it might not be particularly pleasant – nor efficient.

Thanks, however, to the development of AI, we can imagine a learning tool that helps with the creation of on-the-fly immersion content that is responsive to a learner’s need.

The basic setup is as follows. A learner begins a course of study, and then is given some default material to learn from. This learning takes place entirely through audio, and the learner learns by tuning in to the relevant livestream.

The learner’s first task is, simply, to listen. Perhaps it would be a bit repetitive, maybe boring. But simple. One would hope there would be just enough variety to make it interesting. Certainly, that could be arranged. There would be people with know-how to do so.

At the end of some initial period, though, the learner would be called upon to produce something – a short written work, perhaps, or their own spoken work. It could even be a picture or a diagram. The material then would get ingested by AI and this would produce derivitive works. Commentary, perhaps, corrections. Highlights that present the best of learner’s work. Paralell material that helps the learner take their thinking further. All of this text would get piped back into text-to-speech and the radio station would begin to change.

Imagine a learner studying Latin. She begins by listening to verb paradigms, to simple sentences in the target language. Perhaps they would be accompanied by English translation, but the material would be simple and repetitive. At night, however, she practices writing out the paradigms, copies some passages from a textbook, attempts to make her own composition. The AI could review this performance and generate new materials – a new set of paradigms to learn, variations on the learner’s most complicated sentences. A new short story is unlocked for the learner to listen to throughout the day. The more the learner practices, the more variety is created, and the more the material reflects the learner’s interests and passions.

Another learner is studying geology. The programming begins with simple programming about the geological epochs of the earth, the types of rocks, and an introduction to some elements from the Periodic Table. The learner starts taking pictures of the different geological features around his home, and the AI can then create from this a running lexicon of useful terms, of rocks studied, or it could go deep into how these nearby formations came to be and interesected with geological epochs encountered during the default stream.

Indeed, if the learner’s primary complaint falls upon the monotony of the livestream, learner activity would cause it to change more rapidly. In this sense, it is different from other input-focused language learning programs. Output becomes as important – if not more important – than input.

It may even be the case that the urge to affect or alter the programming on this radio station becomes a primary factor in engaging actively with their learning, just as someone standing before a mirror may be motivated to lift his arm, examine it, or look more closely upon the side of his face.

The classic action of learning is input, then output. A learner makes herself an echo. Only later does this learner make any mark upon the world.

Here the action is reversed. The echo is outside the learner’s ears. It sits atop his heads and speaks.

And just as in the story of Narcissus, where Echo attends upon the one who looks into the mirror, she oversees the learner who looks upon himself – only in this myth, it’s not out of toxic self love, but self reckoning. The first Narcissus became a flower, but the new Narcissus outgrows the water. He starts to become a lover, not of himself, but of everything, the great amateur.

CONNECT. Join the 24/7 Universal Writing Session.

To get updates about this project, sign up to the Substack below.

"Protoclassic helped me both refine my writing and thinking on substantial intellectual projects all the while keeping an eye to the rapidly evolving online scene. It helped refine and then amplify my message while significantly broadening audience reach and impact." — Nicholas Gruen