Mind and Iron: The Future of Content (Is Bizarre and Interactive)
What a Paramount presentation tells us about where we're headed. Also, GPT-5 arrives
Hi and welcome to another spicy episode of Mind and Iron. I’m Steven Zeitchik, veteran of The Washington Post and Los Angeles Times, senior editor of tech and politics at The Hollywood Reporter and reigning ring-ross champion of this newsy state fair.
Every Thursday we come at you with the future-world news you need to know and the ways we might build said world better than we’ve been doing a lot of other things. Please consider joining our community.
On Thursday afternoon, as Sam Altman was presenting GPT-5 on a livestream, I was in a midtown Manhattan office tower listening/talking to David Ellison and his executive team about the future of Paramount. The son of Oracle founder Larry Ellison, David has just taken over the 113-year old studio and was outlining to reporters his tech-y plans to bring it into this decade.
At first these seemed like very different events, but I soon realized they were in fact pretty similar, going right to the question of what our lives will look like on screens in the years ahead. Which is to say, what our lives will look like in the years ahead. So we'll try to make sense of this landscape as it's taking shape before our eyes.
Also this week, I wanted to call your attention to a piece just posted by my former LAT colleague Matt Pearce, who writes on his Substack about the new tech age's problems of not enough consumers. I know, shopping is everywhere, capitalism is amok, how could this be. But Matt lays out a very compelling case for why the ability to produce stuff is about to seriously outstrip our ability to consume stuff (it already does but it's about to really go off), and what that will mean not for companies but for us and our lives.
First, the future-world quote of the week:
“You're going to live in a world where it's entirely possible for a character you love to have a conversation with you on an application."
—David Ellison, new owner of Paramount and son/scion of Larry Ellison, on what our media will soon look like
Let's get to the messy business of building the future.
IronSupplement
Everything you do — and don’t — need to know in future-world this week
What GPT-5 and Paramount’s reinvention have in common; Will we soon be producing things for AI?
1. RIGHT NOW, THE GAP BETWEEN THE MEDIA WE CONSUME AND THE MEDIA WE CREATE is pretty clear. For most of us, we've got all the material we generate in our day-to-day lives — that email we write, that voice note we send, maybe the occasional birthday card we design (with the help of an image generator, or all by our lonesomes). Then on the other hand we have the media we consume — the news outlet we read, the podcast we listen to, the streaming show we binge.
Social media is of course the nexus point of these two lines — we're often just as likely to be dashing off our own post as reading someone else's. But for the most part these vectors stay parallel, never to meet.
What happens, though, when, the two start intersecting in sharp ways? When all the streaming characters become entities we can interact with or even design? When the podcast isn't a host or celebrity talking to us but us in dialogue with them?
On Thursday Paramount's new owner David Ellison, son of world's third-richest man Larry, sat down alongside his financiers and executives to talk about the future of their company and media in general. It will look, he said, pretty much the way I just described.
Amid all the buzzwords about being a media company that's also a tech company (read: why are TikTok and YouTube eating consumer mindshare/our lunch?) and all the pledges to improve streaming services and interfaces and algorithms and all the features that would essentially make Paramount a better Netflix, came the idea that something far more ambitious is in the works. Something that would turn Hollywood content into content social-media malleable. Something that would turn traditional entertainment into a beast most unrecognizable. Something that would erase lines between creation and consumption.
After all, most tech companies (even Netflix, and certainly Apple, Google and Meta) don't own the characters that have been entertaining us for decades. Paramount is one of the few that does, SpongeBob and Ethan Hunt, Snooki and Captain Kirk. So the idea would be, well:
"Once you start actually getting into the world of real-time-driven and powered by frontier models, you're going to live in a world where it's entirely possible for a character you love to have a conversation with you on an application. You could have a ten minute dialogue with it, " Ellison said.
That's as clear a description as we got. And it actually was pretty clear. Most of the language at this presentation and others of its ilk is of the C-suite-speak variety, about "unlocking value" and "mining IP." (You can check out the insightful decodings of my THR colleagues Alex Weprin and Erik Hayden here and here.) But the above comment was meaningful for those of us who've never gotten near an MBA program: it means that a media company's creations will no longer be static, gatekept entities. These characters that began as a human being's abstract dream; were put out into the world via production and then marketing; were codified by our legal system; were emulsified by our culture into something living if not sentient — these characters are now taking the next step of actually coming out to interact with us.
So the characters that drift across our screens whenever a marketing executive (or algorithm) decides they should will now become something that we actively decide when to play with it and how to play with it, a kind of action-figure come to digital life. You'll, say, send a video of Tom Cruise hanging off a plane to your friend while they're stuck in a two-hour delay on the tarmac; you'll talk back to Kyle and Cartman for ten minutes (and they'll answer) when you need a little more ribaldry after the night's "South Park" episode; and yes, you'll eventually design your own ending to “Star Trek” or your own new adventure for SpongeBob. (Or with many other properties not owned by Paramount.)
When people describe this kind of world (it's also happening with an Amazon-invested app called Showrunner) they tend talk about it in terms we already know — as a leveling up of meme culture. But that's more true in form than substance. Right now we use a meme after a company created it; at most when we pull it off the rack to drop it in to some new context we're curating it. But in this new Ellisonian world what we’re doing is creating — and pushing beyond just a three-second expression of an emotion to instead tell new jokes we've conceived of, new vignettes or ideas or even longform stories we’ve imagined. (Well technically it's not new, since models are inherently based on what came before. But the output will feel new.)
There are some obvious consequences to all this. All those gripey fan-board posts now go away, for one thing. You don't like what a TV writer does? Do it yourself. But to look only at such effects is to miss a much larger overhauling of a whole ecosystem. Because endings are not the only variable that's been changed. By allowing interactions with characters in this way we've shifted the whole model, away from any studio or company (and thus potentially any creator that gets paid by them) committing to something original. Why take a flyer on a whole new story when you have an existing story that people love and know? This is already the logic of reboots, but even reboots involve some new story or some extension beyond the original, because otherwise why watch it. In this new universe the sequel isn't what a note-addled screenwriter decides. It's whatever we want it to be.
Let a thousand personalized fragments bloom. Or a thousand pieces of slop splatter, your choice of framing.
Once we take all of these movies and TV shows and games and books and podcasts and let people get in there and play around, what we've done is not just make one small tweak at the end of the process — we haven't just handed Doc Brown the letter about 1985. We've ripped a hole in the entire time-space continuum.
None of this is to say there won’t be reason for professionals to tell stories. Certainly they’ll want to, innately. But the level of financial risk for the studios that hire them, already so high, just got even higher — Skydance-high, you could say — given how profitable and cheap it will be to work in these personalizations. And these studios, already so risk averse to give us something new and professional, will become, I fear, even more tentative, almost to the point that outside of a very select few breakthroughs, we won't see much of it at all. Could a new indie world blossom to fill in these gaps? Maybe. But given how few points of distribution there are and how much they're controlled by the biggies, I'm not sure they'll be meaningfully accessible.
Tl; dr — this new world will gives us an incredible ability to make all our favorite characters come to life and do things we (or their creators) never dreamed of; the world will feel much more possible and fresh than when we just, pity the naivete, played with action figures or wrote fanfic. We just won't get many new favorite characters. And that’s a pretty damn consequential potential outcome — the Dream Factory, that quintessential American facility, sitting ghostlike and gathering cobwebs.
How soon is now on all this? We're a long way and not a long way. The long way because the models until now can still be pretty stiff when it comes to sounding like a full-fledged fictional character or “digital twin” of a real person, susceptible to hallucinations/contradictions and plenty prone to recursive loops. Like the hotel room in "Oldboy," the walls of possibility may be more closed-in than we see. And that's just for audio or textual interactions; video is a whole other ballgame. The idea that we can just custom-order ten minutes of video and have it feel like even a rough approximation of “SpongeBob” or “Star Trek” has, until now, been a big climb indeed.
And yet. There is an entire generation that sees content not as something you passively get served but something you actively create. There is, at the other end of the food chain, an incredible incentive for companies to deliver these kinds of modalities. And now with the Paramount ownership change we have a company — really the first company — that own these decades-old properties while also having the money and nimbleness to do something like this with them. (Netflix and YouTube, popular as they are, just aren't sitting on the level of belovedness.)
So the question is the models: Can they generate conversation and ideas and jokes and scenes and videos so naturally that it will allow Ellison’s ambitions to become the norm. Which brings me to the GPT-5 announcement.
As you likely heard, OpenAI released its new model on Thursday to the general public, deciding they can split the atom on GPT-4, GPT-4.5, GPT-4o and other alphanumeric combinations no more. On many levels GPT-5 is not what was long rumored/supposed to be. (I mean, it was long rumored/supposed to be AGI, Artificial General Intelligence, and even Sam Altman admits that ain't happening anytime soon.)
GPT-5 is instead mainly an incremental improvement on what came before — it "thinks" (man I hate that verb) longer so its answers to tricky questions are more comprehensive; it can decide which of its models new or old to send queries too so its easy answers come back faster (in fact this its main M.O. — a kind of AI-model traffic cop); and it generally operates at a level that increases the accuracy of its outputs, at least according to the early indications. GPT-5 is now the default model for ChatGPT even for free users so everyone will be able to experience for themselves what it can/can't do. The more pro upgrades are pretty cheap compared to their competitors too.
What's a lot murkier is whether it really offers qualitative improvements. "We are introducing [XYZ,] our best AI system yet," is a line we've heard from OpenAI before, and we do here. And it's true, relatively.
But the company continues with a bigger boast, that GPT-5 is "a significant leap in intelligence over all our previous models, featuring state-of-the-art performance across coding, math, writing, health, visual perception, and more." That part is less clear (the early indications are that the writing isn't that, for instance).
So far the reasoning on the tricky problems seems solid enough. And GPT-5 does at least show some flair on the creative side , coming up with a little more cleverness or poetry in turns of phrase that doesn’t descend into the soulless-robot valley quite as much, as the Substacker and academic Ethan Mollick notes. But as Mollick also notes, GPT-5’s main advantage is that it allows even free users to access its deeper reasoning models — that is, an advantage of consumer interface, not high-level model operability. Altman himself conceded this, in his way.
"GPT-5 is a major upgrade," he told reporters . "It’s a significant step along the path to AGI. [But] this is not a model that continuously learns as it’s deployed from the new things it finds, which is something that, to me, feels like it should be part of AGI."
That continuous learning is also something you need if you want AI SpongeBob to sound like a dynamic character a screenwriter is writing in front of you, not just a rehash of a bunch of stuff he’s said before meant to sound new. Given Thursday’s developments it’s not clear we have the tech to make that happen yet. But also given Thursday’s developments we have a very rich and powerful person who controls SpongeBob now devoted to ensuring it does.
2. SOMETIMES AN ESSAY COMES ALONG THAT’S SO PERFECT, AND ALSO SO EYE-OPENING, THAT YOU REALLY CAN’T DO MUCH MORE THAN REPRODUCE IT.
So it goes with Matt Pearce’s new essay “The catastrophe of knowledge work waits to be beautiful again, and interesting, and modern,” published this week on his Substack. Pearce is a longtime journalist-turned-labor-advocate (for a free and human press), and the novel case he’s making here is that we’ve gotten so good at producing stuff — from widgets to academic papers — that it’s kind of lapped our ability to consume it.
This is true on a purely metric level — the amount of research papers published has increased notably in recent years, for instance, even as the number of people reading it has stayed flat. And it’s true on a sensory level: how many times have you gazed upon some new creation, content or otherwise, and thought ‘I wish I had the time.’
The even more novel case is that AI can solve this problem, at least on the information-product side, if solve isn’t doing a little too much work in that sentence. Bots and crawlers and summarizers can hoover up everything being published, posted and distributed (often by a hoovering AI itself) and then filter and relay it to us. Executives don’t read email; citizens don’t read journalism. AI does, and passes it along. It becomes the primary consumer, and maybe to the good, as it can actually handle what’s being sold.
“One obvious way to juice consumption is make robots do it,” Pearce writes. “There are hard biological limits, after all, to humans’ attention spans and reading speeds.” (By the way this does not contradict the Ellisonian view: no one said that what’s being created is necessarily new.)
Pearce admirably refrains from making too much of a value judgment on his proposition — after all, it’s dystopic, but maybe also kind of necessary if we’re to keep up? As he writes, “It’s not a particularly idealistic situation. But it could be living.”
But whatever you make of it I suspect he is getting at something that will define our lives in the information-blizzard, content-avalanche world of the late 2020’s and 2030’s — a feeling that the AI-ification of content production will make the level of overwhelm ten times more unmanageable than it is now, and a sneaky but not crazy solution is that machine intelligence can help make order of it. Anyway, check it out. Before our AI Agent starts doing it for us.
The Mind and Iron Totally Scientific Apocalypse Score
Every week we bring you the TSAS — the TOTALLY SCIENTIFIC APOCALYPSE SCORE (tm). It’s a barometer of the biggest future-world news of the week, from a sink-to-our-doom -5 or -6 to a life-is-great +5 or +6 the other way. This year started on kind of a good note. But it’s been fairly rough since. This week? Not too awful.
3D AI ACTION FIGURES, BUT WHERE’S THE NEW? -2.5
GPT-5 IS MAINLY AN INCREMENTAL IMPROVEMENT: -1.0
UNLOOSE THE ROBOT READERS: Uncomfortable but slightly helpful? +0.5