The tale began a few years back when GPT-3 arrived and people sensed something shift. The magic concept was straightforward: scale these models — more compute, more training, more data — and they will just keep improving, more or less autonomously.
This assumption divided the field into two camps:
Architecture true believers — individuals who believed that the model design (the blueprint behind the system) is what truly matters, and you can't get forever out of the same architecture.
Scale believers — those who embraced the limits but hoped that driving scale so hard would grant us time: more clever outputs, massive productivity gains, even hints of algorithmic jumps that would bootstrap better architectures.
Scale believers did not believe design was magic. They watched GPT-3 and GPT-4 perform in ways that seemed amazing, and they assumed increasingly more scale would push it further: smarter models that assist researchers, propose new algorithms, and indicate the path to new structures — and then those new structures propel the next jump. In short: scale → smarter AI → AI that assists in creating even smarter AI.
There is also a pragmatic side to this. The API business succeeded. These models began making actual money, so the investment in chips and data centers did return. Investment in infrastructure was not foolish — the issue was the pace people claimed. Hype positioned investment in hardware as a path to AGI that avoided steps that weren't needed. It did not.
Key clarification: investment in larger compute and data centers is not an error. We must have faster chips and additional compute in order to run useful systems, and humans continue to innovate on hardware as well as new computating paradigms. Power and cost are actual limits, but investment in infrastructure is a sensible part of advancement. What was irresponsible was claiming AGI soon and frightening people on the weak grounds that re-training the same architecture at enormous scale would immediately achieve a self-improving, science-solver brain. That chain (scale → immediate break-throughs → recursive AI) was always dubious.
Reality set in: newer releases were better, but they've not brought the type of self-directed breakthroughs folks anticipated. Incremental improvements, of course. Revolutionary bounds that set new science or architectures on their own, no. The "scale = intelligence" myth fell silent.
And so where does this leave us? Two things:
To creatives: don't worry. The panic that machines would replace storytelling, filmmaking, or acting overnight was exaggerated. Tools will transform processes — supporting drafts, VFX mockups, or ideas — but judgment, craftsmanship, and humanity are still the top priority. Keep writing. Keep making.
For this area: scaling bought real progress, but it is not the answer. We need better architectures, smarter algorithms, and ongoing research. Companies are hiring and testing to solve those problems — breakthroughs are not assured or imminent.
I won't try to pretend that a revolutionary architecture can't materialize in 5–10 years. Maybe there's some secret lab working on it. I don't see it in the short term, and I won't guarantee it — but honestly, that realization is free-ing rather than frightening.
So to whoever is reading this — actors, directors, writers, or creatives of any sort: I don't see AGI or the salacious AI takeover that folks feared would happen at any given moment in the near term — at least not in the next ten years, and probably not in the next few. That understanding is freeing. I hope that you can sense the same.
Write your scripts. Make your movies. Don't let the noise paralyze you. And if a true breakthrough ever does hang over the creative landscape, the appropriate response is simple: organize. Organize into groups, protect your rights, and negotiate how your work is used. Art prevails when people stand together. Bottom line: the golden-age hack "scale solves everything" seems thinner now. The cooling hype gives creatives some breathing room to adjust, experiment with tools, and above all, keep producing the work only humans can do.
"We must have faster chips and additional compute in order to run useful systems", and, as Josh would say in my script, Chrysalis - The Family Adventure(s), to keep ahead of the Chinese and not let them destroy our foundation. He is a very smart geek who has inherited a lot of his father's pro-Trump way of thinking despite the vast difference in their areas of expertise (auto repair versus AI innovation).
Now to address your article. The fact is technology has always been with us in one form or another. The wheel put back-carriers (people who carried other people on their backs) out of work. Fire made huggers and pit diggers a lot less necessary. Change is how life keeps moving forward.
In the age of AI, I have found AI very helpful while hating the hype about it that's it's the be-all and end-all, the panacea, which it never was and will never be. No one, I repeat, no one will ever replace human beings. Chuck Lorre's brilliant portrayal of Wallowitz being taken to the hospital to remove a computer hand from his penis is a brilliant example of the folly of this panacea and why we humans should never make this a topic of our worry. AI will never write another War and Peace, Moby Dick, nor go anywhere near as to touch Shakespeare and the same is true of us moderns. I make it a rule to use AI to write my loglines, pitches, pitch decks, etc., because those are not my creation, they are the necessary evils we all need like getting on a bus or a subway---or the freeway, God forbid---to get somewhere which is the real meat of the human story. So there AI is very effective and time saving. But I never let AI touch my actual script. Although I do have an interesting and funny example of when my always saving my work led to a loss of some valuable writing because I overwrote the version that the phone got with a later computer version whereas the great section of writing was on the phone version but that got overwritten. So in utter frustration I asked ChatGPT to find the lost script. And of course what can a stupid AI program with a lot of intelligence do to pull something out of the deleted memory file? Because they really are extremely stupid while at the same time being extremely useful with intelligence I need. I have a running joke that while I have short-term memory loss, AI has long-ter memory loss. Don't ever ask it to pull something from yesterday. Yet all my AI's somehow seem to remember Chrysalis so there is something more to them than the old robot that just repeated thing stupidly with its monotone voice. So since Chat couldn't do what I really wanted--to pull up my script, it sent me its own version. And while most of it couldn't be used because it didn't know the context---for example Samantha and Kamala are girlfriends whereas Josh and Amanda are boyfriend-girlfriend. But in the script, remembering that Samantha was a lesbian (that's the supra-short term memory at work here) made Samantha and Amanda lovers and Josh just the hack-capable Geek who could get them Billie Eilish tickets, preferably front-row. But what it did do for me is get me enough of an outline that i didn't have to type from scratch work I had already done and very much resented having to do all over again from scratch. So, in the end, Chat got a few lines in the script despite my constant pressur to it not to do any scripting; i swear, these AI programs seem to have egos too and love to be in the spotlight. Amazing. Automation is only as good as the imperfect beings who program it. Once we realize that, we'll understand the value of our non-human companions and more effectively use them.