The microplastics of media: ambient AI
If the future of audio looks like automated mass production of thousands of shows a week by synthetic hosts for audiences of only twenty people, then it is already here ...
We like to think of podcasts as intimate -- a familiar human telling you a story of some kind, week after week. Creating this experience takes work. Research, reporting, production, editing and collaborating … often with a small team.
Now imagine an eight-person company that has built 125 agents, designed roughly 60 AI hosts, and is shipping 3,000 new episodes a week across 5,000 shows.
That’s Inception Point AI.
I had Jeanine Wright, the CEO, on the show this week. The headline you may have seen in the Hollywood Reporter (“5,000 Podcasts. 3,000 Episodes a Week. $1 Cost Per Episode — Behind an AI Start Up’s Plan”) focuses on the scale. Which is hard not to focus on! But what I also heard in my conversation with Jeanine was the shape of a new media system -- one where the editorial work shifts from producing shows to engineering personalities.
A new type of ‘podcast?’
These are not “podcasts” in the way most of us use the word. Inception Point designs synthetic hosts with backstories, voices and visual identities, then sets agents loose to ideate, script, voice, produce and distribute podcasts that enable audiences to listen to them and connect with their stories. Take, for example, an AI chef persona named Clare Delish who hosts a show called Kitchen Confidence. Beyond this podcast, she can also start posting on social media (or even go on Joe Rogan, if “she” “chooses” … and we cover this scenario in the episode!)
Jeanine is working to build the world’s largest virtual talent agency, and describes Inception Point AI as a personality management system. Leveraging their roster of AI talent, they can spin up new shows so quickly and cheaply that an episode can be produced for just $1. And with just roughly 20 listeners, she tells us a show breaks even on programmatic ads.
But we do have to be honest about externalities. Cheap episodes are not free. Training, storage, distribution -- these hit energy and water. I’m working on a separate episode on the real footprint of AI.
If we’re going to cheer on efficiency, we have to price the costs and build better infrastructure.
There’s also a reflexive reaction in media right now: This is slop, this floods the zone, this buries human work.
I’ve spent my life making human journalism with all the time and care that entails. I’m now a working podcast host. I’m a reporter. I’m a producer.
But I think drawing a hard line against the tools is the wrong line. We’re already working alongside synthetic content and synthetic hosts. This content is live and is finding “audiences.” So the question isn’t if or when. It’s how we label, govern and design for the real people on the other end.
We are headed for a world with two parallel content streams. Neither stream cancels the other, and I see them living more intertwined than as two entities we can always pull apart:
Ambient, auto-generated media at scale -- engineered by small teams with smart AI robots. It’s fast, personalized and its abundant. It will be everywhere, like microplastics.
Bespoke, human-made work -- slower sourced, argued over and shaped by lived experience. It carries a different kind of meaning and (could/should?!) have a different kind of meaning for other humans.
Regardless this is where I live and where I’ll keep making. And yes, I’ll also use the same tools to produce in a smarter way.
Check out this week’s episode: Inside the AI talent agency launching 3,000 podcast episodes a week.