Hopefully you’re here after reading my review of The AI Doc, or How I Became an Apocaloptimist. If you did, you know I’m not a fan. The AI Doc purports to explore the dangers of AI (before assuaging them), but does so by lobbing softball questions at CEOs who have a direct financial stake tied to the practice of what has been called “criti-hype,” which is handwringing about, and then handwaving away, hypothetical problems… that might be caused at some point… by feeding word guessing machines even more stolen human output.
While these AI bros - both those in favor and those (allegedly) against - spend their days “yes and-ing” the doomsday scenarios once a browser tab running a chatbot wakes up on the wrong side of bed, these men conveniently memory hole the very real AI harms of the right now. That seems to be because they’re too preoccupied with monetizing those AI harms. These include the connections to global authoritarian power, the deep roots in the racism of eugenics, the overt misogyny, the worker exploitation of the Global South, along with the idea that the AI datacenter coming to a town near you is as inevitable as the AI bros claim it is.
Filmmaker Valerie Veatch’s riveting Ghost in the Machine refuses to ignore these urgent matters of the here and now. As you might expect, it also doesn’t mention The AI Doc during its 90-minute runtime, but if you’ve seen both, it’s hard not to measure one against the other. By giving voice (and space) to individuals and perspectives that were omitted from The AI Doc, Veatch’s documentary takes a poleaxe to the indulgent solipsism of that film.
(You hear that same solipsism in their every media appearance by the way, including in this latest news cycle’s empty overtures about “pacing the frontier.” Please help us, we poor, hapless CEOs. We know what we’re making is going to kill everyone, that is if it doesn’t make everyone useless and want to kill themselves before our product does, but we can’t possibly stop creating this thing, as the thing we’re making is inevitable So yes, we need more capital and regulatory capture and a lot of land and water outside of Des Moines, but with your help, we’ll take our foot off the accelerator just a bit. Because China!)
But back to the film review, and once again to using one film to define the other. Variety’s Owen Glieberman said of The AI Doc: “If you have any interest in artificial intelligence (which is to say: the future), you should go out and see it right now.” Contained in that sentence is one of the most glaring, and I believe most dangerous, problems surrounding AI news and analysis (I’d put Glieberman’s movie review in the analysis category). It’s framing AI in the future tense.
The problem with this framing is that almost no predictions from either AI boomers nor doomers have come to pass. Chatbots haven’t solved human mortality, reversed global warming, emptied my dishwasher, and the last time I went out to eat, there were plenty of humans in the restaurant making and serving the food. Meanwhile, I’m still flesh and blood wrestling with a keyboard; AI has not yet figured out how to use my biomass to power a GPU.
So if you’re a layperson with a genuine interest in Artificial Intelligence, it seems that you have a choice about where to direct that interest: you can either live in a fantasy world of doomsday and/or immortality, or you can join real humans in the real world and take stock of what AI can actually do, what it is actually doing, and most importantly, who AI is doing it to. In other words, the AI bros keep spinning tales about the future. Ghost in the Machine invites viewers to pay more attention to the present, and also to the past.
With that, here are just a few of the film’s most interesting revelations/observations about that present and past, in no particular order:
AI is, and has always been, a marketing term.
The phrase “artificial intelligence” has always been a marketing/fundraising term. Ghost in the Machine even dug up some remarkable footage from the famous 1973 Lighthill debate, held at the Royal Institution in London. The debate featured American computer scientist John McCarthy, who told a packed house in London that “...I invented the term Artificial Intelligence because I was trying to get money for a summer study.”
Linguist Dr. Emily Bender later adds:
“Artificial Intelligence is just a marketing term. It doesn’t refer to a coherent set of technologies.”
As I covered in the book Lies My Tech Bro Told Me, McCarthy was referring to a 1956 Dartmouth summer research project he established, an event that’s now viewed as the moment of conception for this specialized field of computer science. By the time he had moved from Dartmouth to Stanford - Stanford being an incubator of right-wing political thought - McCarthy politically transitioned from Marxist to, in his words, right-wing republican. Also while at Stanford, McCarty published a tract called “Technology and the Position of Women.”
As with David Sacks and Peter Thiel who also pickled their brains with misogynist thinking while at Stanford, McCarthy was concerned that too many women and people of color were being permitted to walk the hallowed grounds of Stanford.
Which leads me to another takeaway from Ghost.
The racism and misogyny of AI is a feature, not a bug.
Building on the foundation that the phrase “artificial intelligence” is an empty, 70 year-old marketing term, the film transitions to an examination of the overarching goal of AI, which is artificial “general” intelligence, or AGI. There is no agreed-to definition of what AGI is in the first place, but that isn’t stopping frontier model labs like OpenAI, xAi, and Anthropic from heedlessly pursuing this goal.
That goal, as OpenAI says in the OpenAI Charter, is creating a “highly autonomous system that outperforms humans at most economically valuable work.” In other words, a computer that is smarter in all ways than a human. (And is also conscious, which in turn would make a slave. But that’s really beside the point, unless of course it isn’t.)
Again, from Emily Bender:
“The acronym AGI…is basically a kind of hype inflation. So when Artificial Intelligence got over-applied to too many things, and people still wanted to be selling this idea of an autonomous thinking machine, they had to come up with a new name for what comes next… but anytime someone is comparing their computer system to what people can do, they are presupposing this ranking (of intelligence, like IQ does)... it’s incredibly dehumanizing and incredibly problematic.”
The fly in the AGI ointment, though, is wrestling over how we collectively decide what’s considered “intelligence,” and moreover, who gets to decide. Spoiler alert, the how is with tests and algorithms, and the who is racists, or at the very least the racist-adjacent. After all, who might you guess tends to score well on tests written by middle aged white men of Anglo-Saxon heritage?
Ghost later discusses how Silicon Valley has created the “great man” mythology around the “innovators” associated with the Internet, social media, and now what is being called AI. Things that would have been invented anyway - using the Internet to send pics, or sell books, or to connect with friends, for example - get recast as acts of unique genius from the minds of unique men like Marc Andreesson, Jeff Bezos, or Mark Zuckerberg.
But in that mythology, along with the power and wealth the mythology helps create, become self-perpetuating. In this environment, the priorities or considerations of women or people of color take a back seat to the interests of wealthy white men (and their hangers on). In this environment, the mental health of young women becomes a while the Grok chatbot from Elon Musk, when it isn’t creating sexualized images of young girls, is worshiping, in July of 2025 no less, “the greatest European of all time… his Majesty Adolf Hitler.” (I wonder what German political party someone who creates a such a product would support?)
In this environment, the mythologizer and hanger on Lex Friedman asks this of Elon Musk during a podcast:
“...how should (AI) handle cases where the raw data shows differences between different demographics, genders, or classes of people—especially when those truths are considered offensive or socially unacceptable?”
Ghost reveals how tech innovation in the U.S., from the origins of Silicon Valley to the latest AI hype, remains haunted by these core logics, along with deeply ingrained misogyny and white supremacy.
In any case, Ghost weaves together several voices to make this specific case, one of those being Dr. Dan McQuillan, a computer scientist who received a physics degree from London’s Oxford University. He points out that:
“AI has its roots in the beginnings of science, in the beginnings of empire. The most important thing for AI is that it has its roots in eugenics.”
AI is horribly exploitative of workers, especially in the Global South.
Ghost also reminds us in the documentary’s second act, and especially in its Chapter 6, that these large language models - the chatbots that are being marketed as AI - are not magic. They are not sentient, they are not intelligent, and we would do well to keep in mind that AI is built using humans every step of the way.
First, Ghost reminds us that creating large language models rely on harvesting massive troves of data from people. All your email is being sucked up by Google. All your photos and posts are being scraped and sold by Meta. Every blog, book, post, pic, purchase, and video posted online. The data harvesting is constant and all-encompassing.
The point being is that humans, either directly or indirectly, create everything an AI is trained on. When tasked with producing original academic research, AI fails.
Second, creating an LLM relies on a massive amount of human labor cleaning up all that data. This is shit work. Of course, this work doesn’t fall to Silicon Valley software engineers being paid with stock options and kombucha, but rather to desperately poor workers in the Global South. As you can imagine (some don’t have to imagine, of course) this means sifting through the most horrendous words and images that humanity has managed to upload to a server. This work of data “labeling” is destroying the mental health of the people doing it, and they’re doing it for a wage of about $1 per day.
The subtext is obvious. The frontier LLM lab CEOs are trying to create digital slaves, and are doing so using a system that closely resembles colonial slavery from 200 years ago. So keep that in mind next time you ask Copilot or Gemini to summarize your Inbox. (Of course, Microsoft and Google have just set their defaults to do it anyway.)
As a final note in this movie review, I’ll also point out that what makes Ghost such a fantastic interrogation of AI is how it engages with people affected by - or speaking on behalf of people affected by - the tangible, present harms caused by how AI CEOs conduct their business. For example, we hear from Nairobian tech workers Richard Mathenge, Mophat Okinyi, and Kings Korodi, who paint a harrowing picture of how companies like OpenAI, SpaceX and Anthropic exploit desperately poor workers in Kenya. (Why Kenya and other English-speaking African countries? Their colonizers spoke English, and now the data that needs to be labeled requires English fluency.)
As mentioned, Ghost provides space for Dr. Emily Bender to actually make her point, and also gives airtime to Dr. Émile Torres, the co-author of their famous Stochastic Parrots paper examining the nature of AI, and the pair who coined the acronym TESCREAL. We hear from Dr. Milagros Miceli, who leads a research initiative that focuses on the experiences of the data workers who train AI systems, the ones who wake up screaming in the night because of images they label by day. “It tears the veil of what makes you human,” says one such worker.
Meanwhile, Dr. Becca Lewis is provides a very clear distillation of complex history of the Silicon Valley mythos. Dan McQuillan, mentioned previously, provides plain-language context about the underlying tech. (He was also apparently the first to highlight the connection between AI and eugenics to the film’s director, Valerie Veatch.)
In all, it’s a lineup of nearly 40 historians, computer scientists, and human rights activists that tell the story. And notably missing from that roster, in a good way, is the CEO from an AI lab, or anyone whose livelihood is otherwise tied to a fantasy that Silicon Valley is building God out of computer chips.
The central question The AI Doc grapples with was this: does filmmaker Daniel Roher worry too much about AI, or does he not worry enough? (Not worrying enough would itself be worrying!) That is, will Daniel Roher’s anxiety levels over AI ever be manageable?
The central question of Ghost In The Machine is much more interesting, and it’s posed during the film’s closing moments philosopher Dr. Johnathan Flowers: “Why does this need AI?”
I think Veatch deftly invites the viewer to supply their own this when pondering Flowers’ question. Does this… this department, this company, this email draft, this movie review, this surveillance mechanism, this bombing run in Gaza or Iran… need AI?
It’s a great question indeed, from a documentary that spends its time asking very good questions.