"The AI Con" breaks down how AI is a parlor trick forced on us by greedy capitalists
AI is a con — and this expert in languages breaks down why.
AI is crap.
That’s a fact (and to be clear, I’m referring specifically to the large language models, or LLMs, that are colloquially referred to as “AI,” even though the term is actually much more complicated). It’s self-evident to those who aren’t afraid to say the emperor is in fact naked. And this brings us to linguish Emily Bender and sociologist Alex Hanna, who recently co-authored a new book, “The AI Con,” that Silicon Valley emphatically does NOT want you to read.
Why? Simply put, because “The AI Con” explains how all of the talk about so-called “artificial intelligence” revolutionizing our lives is bullshit.
“We call the book ‘The AI Con’ because there's a con happening here and it's happening at many levels,” Bender told Back Seat Socialism. “At the base, you have what's effectively a parlor trick, the systems that everyone's excited about. Right now, the large language models or chatbots are systems designed to do exactly one thing, which is to output plausible looking text. They mimic the way we use language.”
This is why, when people use LLMs like ChatGPT and OpenAI, they receive the illusion of a mind at work when in fact all they’re looking at is a super-advanced calculator spewing out verbiage based on probabilities. The result is that it tricks the gullible into seeing a mind, an intelligence, where none actually exists.
“We use language in just about every sphere of activity,” Bender said. “So we have systems that look like they can do all of these different things, and maybe more cheaply than it would require to pay a person to do it, although that's arguable honestly,” but in fact it is not actually “doing” any of the real, thought-dependent work.
“The way we understand language involves not just unpacking meanings from words, but actually keeping in mind everything we know about the person who wrote the text or uttered the language,” Bender pointed out. “Everything we believe to have in the common ground with that person, and everything we believe about what they believe about their audience, whether or not it was us. Given all of that, we say what must they have been trying to convey by choosing those words and in that order? And when we do that, we are imagining a mind behind the text.”
Hence the “parlor trick” of AI: “The problem is we can't turn that off when we're looking at the output of the synthetic text extruding machines, and so they look like they can do much more than they actually can because we imbue them with so much when we look at their output.”
From that “base level parlor trick,” the massive corporations behind large language models like Grok, Google AI, and DeepSeek have duped millions into believing that they have created a real “artificial intelligence.” Their economic incentive is to convince some businesses “you can save lots of money in your workforce by just paying us instead to run this synthetic text extruding machine” and convince other institutions, like the government, “we can handle your education needs, your healthcare needs, for much less if you just give us the money that would go into those systems to us, [which we] then return to you in the form of this synthetic tax extruding machine.”
The dangers here should be obvious. In these scenarios, businesses and other important entities are laying off employees because they believe so-called AI can do jobs it actually cannot. This both increases unemployment and bakes technological incompetence into systems on which we all depend.
“That is both dangerous for the people who be should be served by those systems [and] it's bad for workers because people get displaced,” Bender said. “The stuff is janky. It doesn't actually do what they promise it's going to do. And so the displaced workers often end up being either themselves replaced by somebody else who's in a very casualized, gig worker position overseeing the output of the machines.”
AI won’t just cost thousands of people their jobs, if not millions. It will also make our society dumber on a collective level because — despite tech tycoons insisting AI is all about “intelligence” — the text produced by AI is, without exception, dumber than the text produced by human beings.
“I am very concerned about the pollution of our information ecosystem,” Bender said. “When people take the output of a synthetic text extruding machine, which they're being told could replace a search engine, this is not just on the people who are using it, but it's very much on the companies selling this.” Instead of documents prepared by people in their original context and then evaluating how well it meets a given need, society will have “gone to this all purpose answer machine and taken its answer without being able to situate it or even know it's provenance. And then if that answer goes out into the world, because the person who provoked the machine to give it to them posts it, the worst in my opinion is when you have journalists or other people who are meant to be providing information you’re supposed to trust [sharing it]. It makes it harder for us to find trustworthy information and harder for us to trust it, when we find it we are now swimming in the soup of synthetic text where our trustworthy sources should have been.”
If there is one book all humans should read, it’s “The AI Con,” and I say that not because I’m particularly interested in the technology. I write this because I’m tired of rich people conning the rest of us with their crap.
This time, they’re not just conning us with crap. They’re forcing the crap down our throats and gaslighting us into believing that our gags of disgust are actually moans of pleasure. Reading “The AI Con” is like listening to a therapist who reveals to you that you’ve been gaslit, and empowers you to be able to say to the world that yes, indeed, the crap is in fact nothing but crap.
Back Seat Socialism
Column by Matthew Rozsa who is a professional journalist for more than 13 years. Currently he is writing a book for Beacon Press, "Neurosocialism," which argues that autistic people like the author struggle under capitalism, and explains how neurosocialism - the distinct anticapitalist perspective one develops by living as a neurodiverse individual - can be an important organizing principle for the left.