For a moment, it sounded as though Sam Altman had finally said the quiet part out loud. "We opened Pandora's Box together," he supposedly declares. "The future's inevitable. Countries will fall. Industries are gonna collapse and we get to be the ones to shepherd people into this new world."
If that were really the chief executive of OpenAI speaking, it would be quite a statement. Perhaps it would be time to pack the essentials, head for the hills and make sure one had plenty of baked beans. There is, however, a catch. Sam Altman did not say those words. Andrew Garfield did.
The quotation comes from the trailer for Artificial, Luca Guadagnino's forthcoming film about OpenAI and the extraordinary corporate battle of November 2023, when Altman was fired by his board and returned to the company only days later. Garfield plays Altman, Yura Borisov plays OpenAI co-founder Ilya Sutskever, Monica Barbaro plays Mira Murati, and Ike Barinholtz plays Elon Musk.
Breitbart's description of the trailer makes the distinction rather too easy to miss. The apocalyptic prediction is dialogue delivered by a fictionalised Altman in a Hollywood production, not a confession from the real man. Yet perhaps that is what makes the trailer interesting. Hollywood has taken the diffuse fears surrounding artificial intelligence and condensed them into one terrifying speech. Countries will fall. Industries will collapse. A small group of people building the technology will shepherd humanity into whatever comes next. The quotation is fictional, but the anxiety behind it is not.
Only a few years ago, artificial intelligence was mostly presented to the public as another useful technology. It could recommend films, recognise photographs, translate languages and perhaps eventually drive cars. Much of the initial discussion concerned whether students would use it to cheat on essays or whether office workers could get it to write their emails. The debate has moved a considerable distance since then.
AI systems can now write computer code, analyse scientific literature, manipulate images and video, reason through complex problems and operate increasingly as agents rather than merely answering individual questions. Researchers are experimenting with large numbers of AI agents working together on problems that would previously have required teams of human specialists.
The recent OpenAI claim concerning the Navier–Stokes equations provides an extraordinary example. One of mathematics' great unsolved problems concerns whether these fundamental equations of fluid motion can develop singularities under certain conditions. OpenAI has reported a massive AI-assisted effort involving thousands of agents working in parallel on the problem. This is discussed at another blog piece today.
Whether that work ultimately settles the famous Millennium Prize problem remains under mathematical scrutiny. Terence Tao and other mathematicians have raised important questions about exactly what has been demonstrated. But even if the strongest claim eventually has to be qualified, the episode tells us something remarkable about the direction of travel. We are no longer merely asking AI to explain mathematics. We are asking armies of artificial mathematicians to do mathematics. That is a profound change.
The traditional computer multiplied numbers faster than a human being. Nobody concluded from this that the computer was replacing the mathematician, because calculation and mathematical insight remained clearly different activities. Human beings formulated the problems, devised the concepts and constructed the proofs. That boundary is becoming less distinct.
Suppose 10,000 AI agents can be assigned to a mathematical problem. Some investigate existing literature. Others develop possible approaches. Others attack those approaches looking for errors. Promising lines of reasoning can be expanded while failures are discarded. The entire process can continue day and night without the machines becoming tired, bored or demanding research grants.
The comparison with human intellectual labour becomes uncomfortable. A society cannot manufacture 10,000 first-rate mathematicians overnight. Each requires perhaps twenty years of education and intellectual development. Nor can 10,000 human researchers easily be coordinated on one problem without creating an enormous bureaucracy. AI agents can potentially be multiplied simply by supplying more computing resources.
Today that process is extremely expensive. Tomorrow it may not be. This is where the fictional Altman's warning about collapsing industries becomes more interesting. The first industries affected by automation were expected to involve repetitive physical or clerical labour. Intellectual professions seemed safer because machines could perform calculations but supposedly could not reason creatively. That assumption now looks considerably less secure.
If artificial systems can perform increasingly sophisticated legal analysis, computer programming, engineering, mathematical reasoning, financial analysis and scientific research, the economic effects will not be confined to factory workers or call centres. They may reach precisely those educated professional classes that previously assumed technological change would happen to somebody else.
That does not mean entire professions disappear overnight. History rarely works that way. The arrival of the tractor did not instantly eliminate agricultural employment. Computers did not eliminate accountants. The internet did not eliminate newspapers the morning the first website appeared. Instead, technology changes the amount of human labour required to produce a given result.
One lawyer equipped with advanced AI may eventually perform work that once required five junior lawyers. One engineer may supervise a collection of artificial engineering agents. A scientist may operate something resembling a virtual research institute from a laptop. That can produce enormous increases in productivity, but it can also produce enormous disruption.
The important economic question is who captures the productivity gain. If a company can produce twice as much with half as many employees, society as a whole may become richer while particular workers become unemployed. The benefits are diffuse and long term; the lost salary arrives on Friday. Scale that across millions of jobs and the political implications become obvious.
This is where Artificial may accidentally stumble upon a deeper problem with its fictional phrase about "shepherding" humanity into a new world. Who appointed the shepherds? The most powerful AI systems are being developed by a remarkably small collection of corporations and research laboratories. Their leaders did not win elections giving them authority to redesign civilisation. They acquired that influence because they built technologies that governments, businesses and ordinary people increasingly want to use.
This does not make them villains. Nor does it mean governments would develop AI more wisely. The history of government technological projects provides little reason for such confidence. But it does create an unprecedented concentration of technological power.
A small number of organisations may increasingly determine what kinds of artificial intelligence exist, what they can do, who can access them and what safeguards govern their behaviour. At the same time, governments will inevitably seek access to the same capabilities for military, intelligence, administrative and surveillance purposes. The technology will therefore develop inside an uncomfortable triangle of corporations, states and increasingly powerful artificial systems. Nobody really knows where that ends.
There are optimists who believe advanced AI could produce an extraordinary new period of human prosperity. Artificial scientists might discover drugs, design better materials, improve energy technologies and solve problems that have resisted human researchers for decades. Cheap intellectual labour could become as transformative as cheap mechanical power was during the Industrial Revolution.
There are pessimists who fear mass unemployment, political instability, autonomous weapons, surveillance states and eventually artificial systems that humans can no longer reliably control. Both groups can point to real developments. That is what makes the present moment so unusual. AI optimism and AI pessimism are not necessarily opposites. The technology is potentially dangerous precisely because it may become extraordinarily useful and powerful. A useless artificial intelligence threatens nobody; a system capable of performing a substantial proportion of humanity's intellectual work is another matter entirely.
This is why the Artificial trailer works even though its most frightening quotation is invented. Hollywood has recognised that artificial intelligence is no longer merely another technology story. It is becoming a story about power: who possesses intelligence, who controls it, who benefits when intelligence becomes cheap, what happens to people whose economic value depended upon intellectual abilities that machines can suddenly reproduce, and what happens when the machines begin producing discoveries that their human operators could not have produced themselves.
There is an irony in Hollywood asking these questions. The film industry is itself among those threatened by generative AI. Scripts, voices, music, special effects and eventually substantial portions of filmmaking can increasingly be generated or assisted by machines. The industry making the warning film is therefore staring at the same technological wave that the film warns everyone else about. Perhaps that explains the unusually sinister tone.
The fictional Altman says that Pandora's Box has been opened. The metaphor is overused, but here it has some force. Once sufficiently powerful AI exists, there is probably no realistic route back to a world in which nobody possesses it. If one company stops, another may continue. If one country prohibits development, another country has an incentive to proceed. The competition itself makes restraint difficult.
That does not mean catastrophe is inevitable. Predictions of technological apocalypse have accompanied railways, electricity, nuclear power, computers and biotechnology. Human societies adapt, usually imperfectly, and new technologies frequently create occupations and possibilities that earlier generations could not imagine.
But AI differs from previous technologies in one unsettling respect. Most machines amplified particular human capacities. The steam engine amplified muscle. The telephone amplified communication. The computer amplified calculation. Artificial intelligence aims at something more general. It amplifies intelligence itself.
If that process continues, the most important question may not be whether AI destroys humanity or creates paradise. Those are the dramatic extremes that sell films and generate headlines. The more immediate question is what happens to a civilisation when increasingly large portions of thinking can be performed by machines.
https://www.breitbart.com/entertainment/2026/09/09/artificial-trailer-teases-terrifying-openai-future-countries-will-fall-industries-are-going-to-collapse
https://www.zerohedge.com/news/2026-09-09/we-may-only-get-one-chance-stop-ai