AI Book Club: The AI Con by Emily M. Bender & Alex Hanna
The tagline of this book reads, “How to fight Big Tech’s hype and create the future we want,” and I have never agreed so fast with a book cover.
It’s easy to get swept away by AI hyperbole in 2026. News headlines oscillate from “AI will kill us by the end of the decade” propaganda and “utopia is around the corner” hopecore. Also, mind you, often propaganda.
Authors Dr. Emily M. Bender and Dr. Alex Hanna both write with perspectives grounded in reality, with Bender bringing a cultural understanding as a linguistics academic, and Hanna as the Director of Research at the Distributed AI Research Institute. Here are some parts of the book that really stuck with me.
Synthetic text extruding machines
I really do enjoy linguistics experts’ takes on AI, because language is culture and what makes us so human. LLMs are literally just extremely expensive predictive text tools that mostly mimic language. This particular passage really sums it up for me:
Language models, problematically, have no subjectivity with which to perform intersubjectivity. Despite the frequent claims of AI researchers, these models do not learn “just like children do.” Simply modeling the distribution of words in a text provides no access to meaning, nothing from which to deduce communicative intent. Language models thus represent nothing more than extensive information about what sets of words are similar and what words are likely to appear in what contexts. While this isn’t meaning or understanding, it is enough to produce plausible synthetic text, on just about any topic imaginable which turns out to be quite dangerous: we encounter text that looks just like something a person might have said and reflexively interpret it, through our usual process of imagining a mind behind the text. But there is no mind there, and we need to be conscientious to let go of that imaginary mind we have constructed.
This is why we like to call language models like ChatGPT “synthetic text extruding machines”. Like an industrial plastic process, language corpora are forced through complicated machinery to produce a product that looks like communicative language, but without any intent or thinking mind behind it.
Machines doing shitty things still need humans to do the shitty things
A common optimistic AI belief is that AI will do all the things we don’t want humans to do. We’ve all heard horror stories about Meta outsourcing content checks to human moderators in countries very far away from the US, to the absolute detriment of the moderators’ mental wellbeing.
You’d think that AI could replace that, and it probably will in the future, but in the meantime, many AI tools still use real people to moderate content. A Time Magazine report found that OpenAI was paying subcontractors in Kenya less than $2 a day to filter out offending content, often exposing themselves to CSAM, gore, hate speech and porn.
This has gone a step further, with tech companies now using a method called ‘red-teaming’ where they hire workers to deliberately feed AI models with inappropriate prompts to then check the outputs as a form of QA. All day long, these workers are having to input horrific requests into AI, and deal with the offending outputs when the quality controls don’t flag the content as unsafe.
As Bender and Hanna point out, “When executives are threatening to replace your job with AI tools, they are implicitly threatening to replace you with stolen data and the labour of overworked, traumatised workers making a tiny fraction of your salary.”
Luddite isn’t an insult
The modern definition of a ‘luddite’ is someone who opposes new technology, but Bender and Hanna explain the deeper backstory behind the name. It goes back to late eighteenth century England, when new power looms were replacing artistan weavers. The workers protested by destroying machines, and leaving a note claiming it was the work of one (potentially fictional) Ned Ludd.
Bender and Hanna further explain:
The Luddites were not against technology. Some Luddites, weavers in particular, were into technologies that helped evaluate the quality of their work, for instance, being able to count the number of threads per inch, such that they could fetch a higher price at market. Luddites were instead against technologies of control and coercion, and concerned about the loss of jobs, health and community.
Look no further than the arts for a modern day equivalent. Musicians aren’t against AI that helps clean up poor quality demos, but rather generative AI that has completely scanned their back catalogue of work and allows an AI-generated track to sit at number one on the singles chart.
Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism and Longtermism (and Racism for sure)
Many AI boomers and doomers look to a set of ideologies that go by the acronym TESCREAL, coined by computer scientist Timnit Gebru and philosopher Emile Torres. To define this in more detail:
Transhumanism is an intellectual and scientific movement that says we should use technology to enhance human physical and cognitive abilities to improve the human condition. Absolute Gattaca vibes.
Extropianism is all about continuous human improvement and the use of tech to push our limits. Definitely veering into eugenics territory.
Singularism is a belief that technological singularity via superintelligence will occur.
Cosmism is a belief that human’s destiny is to colonise space. Vomit.
Rationalism is a philosophy that reason is the primary path to knowledge, over emotion, tradition or faith.
Effective altruism is a social movement that believes it’s best to rationalise decisions based on the greater good. These folks would definitely have an answer for the trolley problem.
Longtermism is an ethical framework that views the long term future as the most important. Suffering now is inconsequential to the future.
At a face level, these all sound reasonable. But the more you dig, the more it becomes clear that all of these assume a future that is fairly white and male. It should come as no surprise that people who subscribe to TESCREAL ideologies include famously problematic neocolonialists like Elon Musk, Marc Andreessen, Sam Altman and Peter Thiel.
So, now what?
Bender and Hanna end the book with their guidance on how to survive the great AI con of the mid 2020s.
Firstly, pop the AI bubble by asking thoughtful questions, taking it upon yourself to resist unnecessary AI use in your day-to-day life, join the library and improve your media literacy. We would do well as a society to also enforce existing regulations across business and tech, and shaping them to fit with the AI of 2026. Looking to the future, we should call for transparency, disclosure, accountability, recourse, data rights, data minimisation, labour protections, and finally strategic refusal.
As they write in the final chapter of the book:
Never underestimate the power of saying no. Just as AI hypers say that the technology is inevitable and you need to just shut up and deal with it, you, the reader, can just as well say “absolutely not” and refuse to accept a future which you have had little hand in shaping. Our tech futures are not prefigured, nor are they handed to us from high. Tech futures should be ours to shape and to mold.
You can purchase The AI Con here.
