CRA unveils ‘Audio x Mixed Media Modeling’ session at Heard 2025
LONDON — There’s no shortage of audio material these days! If you’re looking for the number of podcasts, various internet sources estimate this to be between three and five million. Some, of course, might have had only an episode or two, and some might be dormant, but the fact remains that there are a lot of podcasts. Add to this the 44,000 radio stations, a figure made popular through the UNESCO World Radio Day, and there’s a lot of radio. It would take five human lifetimes to consume just one hour of each. Being realistic, we have all learned to be highly selective and transition between our favorite few podcasts and stations. In one sense, we have almost reached what economists call “perfect competition”: without podcast regulation and with online streams, entry and exit are relatively cheap, and — if we leave forensic content analysis aside — each provider offers an identical product. In short, the product consists of information, music and company.
The numbers I have quoted may seem like infinity since there is absolutely no way we can sample them all. However, imagine if listening possibilities multiplied and genuinely reached infinity. Any subject, any time. Well, it’s just happened. A recent innovation from Google Notebook means you can have an instant podcast-style dialog about any subject. Essentially, you feed the Notebook with a relevant bibliography and ask it to create an audio overview. If you wish, you can specify which areas the speakers should concentrate on. Even in the current world of progress with AI, this seemed scarcely believable.
My fear might be that this would be almost too easy for students or anyone else trying to find a shortcut to knowledge
Stress testing the scarcely believable
So, I decided to stress test the system to the maximum by feeding the type of title that would make many a philosophy student feel faint in their examinations: “Generate a podcast comparing the notion of the self in the Danish philosopher Kierkegaard and how it is understood in ancient India.” And because this is the type of thing I know something about, I could feed it with a relevant and extensive bibliography. Could Google Notebook digest a Ph.D. thesis or two, together with some learned articles, and would it make sense? Four or five minutes later, the file was ready — 23 minutes of, hopefully, scholarly material.
Given the rarified subject matter, it starts cheerfully enough:
Voice 1: Hey, everyone, and welcome back for another deep dive.
Voice 2: Yeah.
Voice 1: Today, we’re going to be tackling a pretty complex but fascinating topic.
Voice 2: Oh, yeah.
Voice 1: The concept of self, and we’re going to be looking at it through two very different lenses…
Voice 2: …okay…
Voice 1: …the 19th-century Danish philosopher Søren Kierkegaard…
Voice 2: …okay…
Voice 1: …and the ancient Eastern tradition of Zen Buddhism.
Voice 2: Wow! That’s a…
Voice 1: I know, right?…
Voice 2: …that’s quite a contrast.
Voice 1: It is. It is. We’ve got some academic papers analyzing both, and, trust me, some of this stuff gets pretty dense.
Voice 2: I bet.
Voice 1: But don’t worry, by the end of this deep dive, you’ll not only understand the core ideas about the self — from both these viewpoints — but also see how they surprisingly connect.
I won’t bore you with the rest of this, but you can see that something highly conversational emerges from dense academic writing. Once the two speakers are underway, it is a dialog of equals in reasonably natural voices. And what’s more, they sound friendly despite the density of the subject matter! Indeed, you can listen to it as if it were a podcast, and frankly, the content wasn’t so bad.
There was, however, a rather major technical slip: Zen reached its classical form in China and Japan, and since — for no apparent reason — the “speakers” alighted on Zen, it meant that they were not talking about ancient India. So, the information would have been misleading if you didn’t have that vital fact at your fingertips. Also, I had expected something more philosophical sounding, but our “podcasters” took a more spiritual angle in their discussions.
Proceed with caution
What does this all mean? As my experiment suggests, we effectively have an infinite amount of listening material on any subject that might interest us. However, it needs to be consumed with caution. Another experiment suggested that such outcomes would be better if a single source were uploaded and a preview requested. Nonetheless, it’s impressive. It probably doesn’t need me to point out the dangers.
As with AI in general, and our example in particular, it has slightly misunderstood the question. As an ex-academic, it is exhausting to think of all the scholarly papers I have read in my life, pondering over footnotes, wondering about the reliability of the sources and questioning the arguments, comparing one view with another to form my own opinion. I’m far from certain that this chatty duo, condensing a batch of seriously dense writing into a highly listenable 23 minutes, would have got me to the same destination. My fear might be that this would be almost too easy for students or anyone else trying to find a shortcut to knowledge.
Nonetheless, it must be conceded that getting the main points across in an easy-to-digest form is extraordinarily helpful for a quick, superficial view of a PDF document. Often, if one has spent too much time reading, audio is very welcome. But proceed with caution.
The author was head of Radio at the EBU until 2020, and before that managing editor of one of the BBC’s national stations. He currently advises media organizations.
More stories by Graham Dixon
Engaging the next generation: News for young audiences
