3,000 Podcast Episodes a Week of AI Slop?

A quick post on the first podcast content farm

There’s lots of chatter about Inception Point AI and how they're churning out 3,000 podcast episodes a week of AI Slop (how dare you called it AI Slop you luddites they shout back).

There are plenty of links to the material, and it’s dreadful, of course. Their model is to make it so cheap and surround it with ads that even a small amount of listening per episode is fine. It’s the audio equivalent of those content farms running Google AdSense. All the visits are fly-bys, and it’s the advertisers who lose out. Listeners will clearly realise it’s rubbish.

Really, it’s a self-contained universe of screwing over ad providers. I imagine that means Spreaker, which seems to host it, will soon stop once their advertisers start to complain. I also imagine Apple and Spotify will remove it from their catalogues. The last thing they want is an arms race rewarding those who pump more and more rubbish into their apps.

So I expect it to be gone within the month. This storm will blow over.

But what I find really odd is that you could use AI to do a much better job with just a little care. I sat here, and in 30 minutes I used AI to create a pretty good first draft of a script for a 30-minute podcast episode about a person. I had it analyse similar podcast transcripts, work out style and substance, test some options, and iterate.

The script is actually just the output of the work I put in. The next person I pick, using similar prompts and again working on the output, would probably take about 10 minutes.

I haven’t done the AI-voiced bit, though that tech is improving. If I avoided the cheapest one on the market, I bet I could find something pretty good.

And this is the thing that raises far more questions than the half-arsed AI slop generators above. This is what’s coming, and it’s going to be much harder to dismiss. These sorts of shows will fill a gap for some listeners. If you currently make something similar manually (for want of a better phrase), it will probably cause you problems.

I’m sure your product will be better, but how much better? 50%? 10%? How much value (cash, listeners?) can you squeeze from that difference? This is the true challenge of AI. It’s happily going to create material that works for a decent number of people. It will take some of the time people spend reading, listening, or watching. The challenge will be doing the things that AI can’t generate, or later replicate.

Successful creative endeavour is going to be half using AI tools to drastically change the model of content creation and half doing the things that are exceptionally human.

I’m amazed at the sophistication you can get out of AI if you prompt and iterate. My biggest surprise about the Inception Point AI crowd is what a poor job they’ve done with it.

AOB

My podcast’s back, The Media Club. All MD and no AI. A fun, profane episode, was out on Friday. Go listen. Or watch!

We’re also trying out a new thing. Every Monday we’re emailing out a summary of all of the media podcasts out there. I’m calling it Briefed. Go have a look and sign-up if you fancy it. This does use some AI!