I don’t really buy the narrative that people will be only consuming AI content in the future. Maybe the content will be optimized and edited by AI, but there’s a reason that hosts like Joe Rogan are popular: because people like people. No one wants to listen to two robots having a conversation.
I think instead we’ll see an advantage for content creators that are charismatic and have brands built around their personality.
The killer app for me would be if AI could generate infinite versions of old, discontinued content: TV shows with the original casts and believable storylines, movies by a particular director, music by a band who’s no longer making albums. I don’t like most media these days and end up just re-watching/re-listening to stuff from decades ago.
If I could just generate a new Star Trek TNG episode each week that was 100% convincingly done, I would absolutely pay for that product. Or if I could just generate a new Pink Floyd album or whatever, and it’s convincing enough—that would be great.
I'm not 100% sure if I really want this or only think I do. My big issue with all the bespoke AI-generated media ideas is the loss of being able to fully discuss them with humans who didn't watch the same thing. It feels like we'll just end up with 8 billion different media bubbles.
I guess it could be like "Nothing, Forever" where there's one main feed everyone shares, or they could have AI-powered chatrooms/forums "watching" alongside you, but both feel like missing the potential/point to me.
I'd like for that sort of AI to generate alternate versions of shows to explore the stories where characters made different choices, or things happened a bit differently and how the characters respond. Or generate Bandersnatch-style movies where you can make choices for characters. I know the director for that movie said it was very time consuming to film all the different paths.
I love comments like this where people talk about what they'd pay for. I hope someone is watching this and starting to build a company around it. That's far better than trying to convince someone to use your made up product that nobody asked for.
I do think that 100% will be hard to convince though, simply because you know it isn't true. The human mind will be looking for the trickery, even subconsciously.
> No one wants to listen to two robots having a conversation.
(OP here) for what its worth, Wondercraft actually agrees with you here. it says so right on their landing page + I ask him about it in the interview. However in this case I may actually be more bullish* on AI audio than he is.
*since a lot of HNers seem to be taking a very strong negative reaction to this suggestion as some kind of civilizational collapse, I do want to say that while I may not personally care about having AI Steve Jobs drivel on with AI Joe Rogan for 24 hours, I could see SOME version of this thing doing well among the general population in the future when they get the format and the topic right (eg start with meditation, language teaching, whatever else niche long form audio/video thing that doesnt actually really need human input, then go up from there).
As you said, it’s a lot of speculation and postulation.
It could be easily tested and proven by content creators.
Many content creators have mastered a style or format (BuzzFeed, Mr Beast), and maybe that style or format is at a point where it is a formula that AI can help with parts of, or completely with.
But what if AI was able to make a perfect copy of Joe Rogan and anyone else and Joe Rogan decided to generate all his podcasts with AI without telling anyone? You could go forever without knowing this :O
This already happened with Lex Friedman. Longer than most podcasts, far fewer original or deep questions per episode, pretty monotone voice that would be easy to get any current TTS engine to emulate, I give it like a 30% chance.
But what if AI was able to make a perfect copy of Joe Rogan and Joe Rogan decided to generate all his podcasts with AI without telling anyone? You could go forever without knowing this :O
> I don’t really buy the narrative they people will be only consuming AI content in the future.
It really depends on what content you're talking about.
Movies: the movie industry is big, glitsy, expensive, and makes a shitload of money. And they like all of that about their industry. The reason the mainstream movies cost so much both to make and to sell is that they are effectively the "luxury car" of media. While they might try and make AI tools a bigger part of workflows (which is one of the reasons SAG-AFTRA is on strike, they already are trying to do this) no company on the level of Fox or Warner Bros. is going to make a movie written entirely by AI, or even worse, an AI generated movie, and sell it as such. It would tarnish the brand.
TV Shows: trickier. Depends on the studio. Lower-tier smaller studios might start leaning on AI tools harder to help them compete with bigger operations, but still, the notion that people would seek out and consume a TV show (or any media really) that was entirely AI generated only makes sense to me if the people are looking to experience AI generated art, for it's own sake. And we MAY get a bubble of that, briefly. But I suspect what AI is going to dominate faster than anything is:
Social media gristle and Influencers: I think this is where we're going to see AI explode. We already have tons of very bizarre and yet seemingly profitable businesses that just run warehouses of video production teams overseas, that churn out all kinds of stuff: sponsored instagram videos, 5-minute-crafts type junk, those weird facebook videos where it's just some dude going "oh just wait till I take this off and it's gonna be amazing" for several minutes. That's what AI is going to dominate really, really quickly. We've already had cases of entire youtubers with multiple channels turning out to be a marketing team manufacturing "genuine feeling" content to promote whatever products, and imagine if you could automate it. You could construct an entire backend that would just accept sponsorships, work them into videos with surface level readings of current social media trends, render it and barf it out onto the various platforms and promote with ad services.
And like the barrier of quality is so fucking low for that stuff, nobody will care if there's weird fingers here and there, minor grammatical errors or even if you use really old school text to speech. Just put a pretty face up there and hey, if whatever character you come up with doesn't catch, screw it. Burn the build, restart with a different one and repeat until you can go viral.
And I think it's extra applicable there precisely because no one chooses that, and yet most people consume at least some level of it because it's so ubiquitous on social media. There is TONS of this stuff and the only way to lower the barriers of entry further to it's production is to make it so you don't actually need a pretty woman in a bikini to sell your crypto coin exchange, or some business casual douche to sell your LinkedIn alternative or market your podcast app or whatever. You just enter the commands, generate the slop, and upload it.
"And like the barrier of quality is so fucking low for that stuff, nobody will care if there's weird fingers here and there, minor grammatical errors or even if you use really old school text to speech".
A while ago, I was looking for something gardening related.
Stumbled onto an article that read like a mix of: written by an Indian person (no offense!) with poor grasp of English, copypasta from Wikipedia or wherever, and repeated 3, 4, 5 times with different wordings.
Whole site was like that. I'd think "link farm" as description wouldn't be too far off the mark.
But you know what? It did show up in search results when searching the web. And while the content was crap, re-worded & redundancies removed it kinda answered what I searched for. Would surely not be the 1st visitor whom this applied to. And probably some tracking info was collected & sold as result of me reading that page.
No doubt there's countless such sites out there. And AI will easily replace most humans involved in its creation.
So I'm with you there: AI will EAT the low-bar-to-entry part of web content for breakfast. Higher up the stack? We'll see.
Oh yeah. Like AI content just barely clears the bar for "good enough" and tons, tons, TONS of the internet absolutely RUNS on "good enough." AI as it exists now is already a perfect and I do mean PERFECT tool if you want to create just reams upon reams of spam. And many websites and services keep their barriers low specifically to attract spammers. And either the spammers or the websites themselves can now generate spam content at hither to only DREAMED OF scales.
Mind you, this is going to make our already barely usable internet utterly INTOLERABLY clogged with repetitive, derivative, bottom of the barrel content... but we've been trending towards that anyway. AI is just going to slam it's foot on the gas pedal. The value of internet advertising is about to crater again and I don't think it ever fully recovered from it's last cratering.
How soon do you think AI will be able to livestream a conversation between 2-5 participants and mimic their personalities, tics, and unpredictable conversational topics so well that they pass as humans?
Edit: let’s assume this becomes possible in a century. Easy solution: a verification system that ensures that the official Joe Rogan Show is authenticated by the real Joe Rogan. No one will care about a computer copy version.
But the real question is how does it compare to the real one, do people listen to it at the same level and for the same amount of time (no). It’s neat, and I work in ML/AI, but it’s missing depth, coherency, and in some way an actual window into the real world.
Because people are capricious and change what they like for no apparent reason. AI would need to re-train on what’s hot and new and that puts it in position of always a step behind.
Second, just the idea of consuming content solely created by AI makes people squirm even if the content is good.
I personally think it is also a pointless endeavor.
"The effects could be generationally skewed as well. Every AI Engineer parent we know of is already hacking together bedtime story generator apps for their kids"
That's just sad.
From my childhood I remember a (big!) book called something like "1001 Tales from the Orient". Think Aladdin & the wonder lamp, and many stories from similar (ancient) sources. There were a few such books around the house. More fairytales than my sister's & my childhood lasted. Great bedtime stuff.
And then people would go for AI generated content? Which most likely draws from books like this. Or Disney stuff, which in turn draws from similar sources. But all mixed up in weird, non-traceable ways.
Why would you feed your kids such crap, when those originals are up for grabs, free or costing pennies?
Fairytales are fine as they are. They don't become better by mixing up in random ways like food colorants in a cake recipe!
As a parent I have looked into this precisely because there is so much autogenerated crap. With some form of AI by my side I can author my own content to a (hopefully and subjectively) higher standard. I can tailor stories to my Childs life events or school curriculum, I can include characters based on real world friends, aquantences, pets etc.
Im currently working on something along the lines of "Detective Rosies & faithful canine sidekick Rufus, the mystery of the broken lamp at grandmas" trying to use this type of stuff to help them feel closer to relations who live far away.
> Every AI Engineer parent we know of is already hacking together bedtime story generator apps for their kids
I was looking for a gift for my toddler niece recently as a 30+ man who has never really looked for kids things before. Going over the shelves at a local shop, I found a book I had forgotten about for probably 25 years now. I teared up a bit thinking about my childhood bedroom and the nightly story time with my mom reading this among other books. When I visited my sister, she asked me to do story time with her daughter and I was clueless but happy to. I asked her to pick out a story to read and she dug into her shelves pulling out more stories I had not thought about in ages and even recognized a couple of them were the worn books from my own shelves, books that our mom had read to my sister first and likely took home from the library she worked at when they had become too worn. The book she settled on was one I was unfamiliar with (Miss Rumphius) but as I read it I realized it was describing a somewhat fictionalized version of a story known to some Mainers about how we ended up with such a population of lupines. It was surreal seeing all the illustrations of coastal Maine towns (I live in one, my sister is far from here) and then realizing the story the book was telling. I look forward to the day she's old enough that I can bring her to lupines growing out in a town like that and make that connection with her.
It pains me to read quotes like the above. To think there will be kids who won't have shared human experiences of bedtime stories because their parents put them on some infinitely generating AI nonsense. Maybe I'm yelling at clouds, maybe I'm a luddite, maybe kids love and benefit from new stories every time. I dunno. I'm no child psychologist but I have noticed kids often like to consume the same thing repeatedly as well, we watched the same episode of Mickey Mouse like 4 times and then spent the rest of the day acting out a specific component of the episode. I'm trying to keep an open mind but a lot of this sounds like destruction of culture in pursuit of "new" but why?
and i'm sure the generation before us bemoans the destruction of culture in the face of youtube and tiktok and instagram and whatnot. my baby sister grew up on teletubbies, which i regarded as vile trash inferior to my sesame street. she turned out more than fine - its quite safe to say she's better socially adjusted than I am.
i think humanity will always have some shared myths and legends that we'll all like to share and i'm certainly hoping that lives on. but there's a lot of room for net new. and you just need to see the delight in a kids' face once - when you incorporate a suggestion by them into a story, or a drawing, or literally whatever - to want to give that joy to them again and again.
theres room for net new, and room for shared history. not zero sum here.
For those only interested in the definition of the new nonsense word:
> "content flippening” — an event horizon when the majority of content you choose to consume is primarily AI generated/augmented rather than primarily human/manually produced
IIRC nuclear tainted steel is less of a concern in modern times. Perhaps AI content will become so normal and people so accustomed to spotting it that today's concern will become a footnote.
The term was originally used to refer to the hypothetical time when the market cap of ethereum flipped to be higher than that of Bitcoin. (It was a little annoying then too).
yes - i want to be clear that i'm not a crypto shill or anything but I do think the term evokes a landmark change that signals the end of a previously unchallenged dominance. hence repurposing it for this. i genuinely listen to HN Recap and PG Essays every single day now and if you asked me a year ago if I'd have any AI generated podcasts as part of my daily rotation I would have laughed at you.
in the meantime - the human podcasting industry is in decline. Spotify is shutting down their stuff (good riddance). NPR laid off a bunch of their people. Rogan and All-In have gone mostly political. etc etc. it is going to be hard to beat machines at staying on topic and probably mostly ad free because of how cheap and consistent they can be.
seems reasonable to draw the lines out and ask when and if AI content takes over... I like thought provoking questions like that, even if, like eth > btc, it never actually comes to pass. It's a thought exercise.
cant help but notice that HN seems to like negative neologisms ("enshittification", "crappening") than positive ones. I would hope for a more technoptimistic, or at least technoneutral, crowd here.
There was a lot of optimism for these things a few years ago. Currently it is the end of the hype cycle, so lot of negativity. These transformer powered generative AI models has been a thing for years now, so most of the initial hype died down a long time ago.
And even though it was late in the hype cycle it was mostly optimism that got upvoted when ChatGPT got released, but since it was so late in the hype cycle of transformer models overall that hype died down in a few months.
hi HN! excited to share this interview with Wondercraft, certainly one of the youngest startups we've had on the pod — but we think you might be interested because their Launch HN went over so well with you all (https://news.ycombinator.com/item?id=37088087) . I also enjoy HN Recap every day and was excited to dig into the behind the scenes. as a content creator I am also very intrigued by the idea that this AI podcaster is beating me at my own game with less effort, and that I'm actually enjoying listening to his podcasts despite them being mostly AI generated.
> As swyx observed, the “content flippening” — an event horizon when the majority of content you choose to consume is primarily AI generated/augmented rather than primarily human/manually produced — has now gone from unthinkable to possible.
This is pure self-promotional advertising copy. Did anyone who thought about this for more than ten seconds believe such an event was "unthinkable"?
"Consuming content" is fundamentally an act of communication. If there isn't another entity with intelligence and intent to communicate with, I'm not really interested.
Although the author gives themself an escape hatch with "augmented" because it seems extremely evident that almost everything will have some small level of "AI" input (arguably it already does... spellcheckers have been around for quite a while.)
I very much believe LMMs are a tremendously important technology with a dizzying array of applications... but generating "content" for human consumption is pretty much last on my list. I already can't read all the books or watch all the excellent movies made by humans!
The real flippening is coming where most people prefer the AI generated content and even AI interactions. You will also. You will learn to love Big Brother.
The main problem with this is that AI can totally subvert expectations, like in the movie “ex machina”. It can simulate as deep of a sincerity as you need and then turn around and screw you. Or may be screwing you the whole time, at scale.
The worst part about AI is swarming. When people discover how to have self-organizing swarms of AI agents, then chaos will erupt.
And the sad part about it is that corporations (a form of AI) can know 100% that will be outcome and still push each other in an arms race to do it in order to grab all that low hanging fruit along the way before their competitors do.
"Consuming content" is fundamentally an act of boredom and addiction to novel stimuli, in increasingly shorter and shorter forms.
Just as the current generation of youth have much shorter attention spans because of growing up with endless scrolling feeds of bullshit, the next generation will consume AI-generated drivel for no reason other than it's novel.
I was being charitable and interpreting it as referring to any type of media usage.
Agreed that the race to the bottom of low-quality stimulus content is a (probably separate) problem, since that trend was well underway long before generative AI.
> Although the author gives themself an escape hatch with "augmented" because it seems extremely evident that almost everything will have some small level of "AI" input (arguably it already does... spellcheckers have been around for quite a while.)
(author here) i guess it could be viewed as an escape hatch, but fwiw I was just directly thinking about the HN Recap podcast, the inciting reason for having this interview in the first place - HN Recap draws its source material from the human writers and commenters on this site, and curation from the voters, but otherwise every word is AI generated. I would consider this a step above autocomplete and spellcheck and that passes my "it's an AI podcast" bar - particularly because there exist prior HN recap podcasts done by humans that did a worse and less consistent job.
I already assume most pseudononymous online commentary and articles are written all or mostly by machines. Usually for marketing or social engineering purposes.
recently every (google) search I made about some JupyterLab settings landed me (top search result) on very similar sounding blog articles on saturn.cloud website -- definitely seemed mass generated (all recent dates, attributed to no author but just "Saturn Cloud", very similar post structure)
Thanks for the interview! All I think about these days is how I would build a moat in the AI space and was excited to learn what Youssef thought. This is a summary of that topic from the interview. Do you guys think this will actually help them build a moat around their product?
---
- Youssef argued their moat is focusing on the specific use case of podcasting and building a product tailored to that, rather than just being a wrapper around underlying TTS APIs.
- While the APIs exist and can serve many use cases, they aren't specifically built to serve podcasting optimally like Wondercraft is.
- By being opinionated in their approach and selecting the best TTS engine for podcasting, Wondercraft aims to provide a better experience than what could be achieved through generic engines alone.
- They are building out templates and defaults tailored to different podcast styles (e.g. interview format) to simplify the process.
- Network effects from high podcast download volumes on platforms can also contribute to the moat by establishing Wondercraft as the go-to tool in the space.
- First-mover advantage in a new niche like AI-generated podcasting gives them time to refine their product before competitors emerge focusing on the same use case.
- Customization options and integration with other tools/workflows could deepen customers' investment in Wondercraft's platform over time.
I think instead we’ll see an advantage for content creators that are charismatic and have brands built around their personality.