First and foremost, I clearly should have got round to publishing this post sooner. Fortunately (or unfortunately), the current news cycle means the books I’ve been meaning to tell you about for a year now are still as relevant today.
Karen Hao’s masterpiece Empire of AI has become an international best-seller since its release last year. Its in-depth takedown of generative AI pioneer OpenAI is just as useful today in terms of understanding Sam Altman’s company, as it teeters on the brink of a trillion-dollar IPO whilst staring into the abyss of its own potential extinction.
Sarah Wynne-Williams’ Careless People, also over a year old, has for example only just come out in the UK, as Meta has launched a legal harassment campaign to try to stop her promoting it. This is notably why she had to stay silent onstage at her Hay festival conference about the book (in a classic “Barbara Streisand effect” moment, this caused the book’s UK sales to top the charts after surging by over 300%…)
Cory Doctorow’s Enshittification theory has not only been proved right countless times before and after his eponymous book came out last year, the prolific Canadian author has even managed to publish an AI-focused follow-up, The Reverse Centaur’s Guide to Life After AI. I haven’t had a chance to read that yet (he writes faster than I can read!) but I LOVE his vision of what will happen once the AI bubble bursts (TL;DR: frugal AI wins!)
Finally, Les Assoiffeurs, by Fabien Benoit and Nicolas Celnik (in French only for now) is a deep and damning investigation into how several different types of companies are currently going all-in to hoover up all the water they can: be they data centres, industrial farmers, or bottled water companies. It’s a damning indiction of the innate clash between profit and planet (and yours truly is in it, but that’s not a reason to buy it!)
Bref, all four of these books, by shining intense spotlights on just how despicable big tech and business can be when looking to maximise profits at all costs, strangely give us considerable hope once we’ve put them down. Not because they’re a cheery read; but because we can only stop evil if we know how evil it really is (and because each book’s critical and commercial success gives that warm, fuzzy feeling that we’re not alone in caring about all this…) So here are just some of their highlights!
Empire of AI: The toxic house that Sam built
For some reason, the story from Karen Hao’s incredible opus – based on interviews with well over 200 people, many of whom from OpenAI – that sticks with me the most is not one the tech journalist broke herself. The New York times revealed in 2024 that OpenAI president Greg Brockman – yes, the one person in tech that’s given the most money to Trump so far – greenlit the transcription of one million hours of YouTube content to train GPT-4. When he was reminded this contravenes YouTube’s terms and conditions, he waved away the objection. In the end he was right: Google said nothing because it was doing the same thing to train Gemini…
This blatant carelessness – cf. Sarah Wynne-Williams – underpins the whole “making of OpenAI” story, which Hao tells so compellingly, the book is almost un-putdownable. Possibly because no other telling of this tale so clearly exposes the foundations of the mess we’re in now: the “bigger is better” lie, the disregard for copyright, the environmental negligence, the doom marketing… alas, it’s all there.
2015: OpenAI is created, (in)famously as a non-profit. Elon Musk and Sam Altman were amongst the co-founders. Back then the two agreed that OpenAI’s raison d’être was to keep AI responsible. Microsoft was an early investor; around the time it threw $10bn into the hat, it laid off 10,000 staff to cut costs. Nice!
When founded, Altman promised OpenAI would “comply with/aggressively support all regulation” and that OpenAI would be the anti-Google, open sourcing the science and being the paragon of transparency, re. Hao. “They would walk back their commitments to openness once the narrative had served its purpose… to avoid bad actors getting their hands on the technology“. Whence Altman’s notorious fall-out with Musk
2017: Chief Scientist Ilya Sutskever “intuitively believed” that the amount of compute would be the key determining factor in attaining AGI (achieving human-level ‘artificial general intelligence’ had already become OpenAI’s real raison d’être). The only ‘proof’ of that then was the ImageNet competition, which had started using GPUs in 2012, and achieved unheard-of levels of image-recognition accuracy by 2017. Colleague Joel Kaplan would later provide the ‘proof’, in his “Scaling is Everything” white paper (2020); and yet OpenAI’s own experience continues to prove that bigger is not better: new, bigger models can hallucinate more than smaller ones. So, this was one of OpenAI’s biggest foundational lies; but still to this day, everyone believes them
2019:
- OpenAI invented GPT-2, predecessor to the model behind ChatGPT (3.5); but didn’t release it, because it claimed to be too scared of the consequences of it getting into bad actors’ hands. The company thereby invented doom marketing, something Anthropic would surpass it at more recently (Dario Amodei made the exactly the same “too dangerous to release” claims about Fable 5 earlier this year). AI remains the first ever sector to attract investment by saying how dangerous its product could be
- Sutskever already knew AI’s growing environmental impact was “undeniable”… but that was fine, because AGI would “counteract the environmental cost specifically”. Whence began, re. Hao, OpenAI’s “huge gamble that you will successfully reach beneficial AGI to counteract global warming before the act of doing so might exacerbate it.” Today, cue the AI sector jumping on gas-powered data centres…
- Emma Strubell’s paper was the first to question AI’s environmental impacts. Google said it overestimated its own impacts by x88 (probably because it was using its own TPUs, not GPUs. They would only share a “median” prompt’s impact six years later…)
2022:
- ChatGPT was relased in November – Sam Altman flippantly released it with just one tweet – almost as an afterthought, as Hao puts it, because OpenAI was so laser-focused on GPT-4 (cf. Brockman’s stealing YouTube content). “How could the company have failed to predict user behaviour and ChatGPT’s popularity so badly?” asks the author. Indeed
- The same year, Microsoft limited-released GPT-4 in India, in direct contradiction of the company’s own rules: the move had not been approved by its DSB (Deployment Safety Board). Altman knew, but omitted to tell his own board. This is one of the elements that led to his failed ousting attempt the following year.
The rest, as they say, is history. To avoid it repeating itself, Hao concludes with a series of recommendations based on redistribution. Of knowledge, so we have more independent oversight (e.g. from scientists and regulators); of planetary resources, so they are better used; and of influence, or education, to deflate big tech-dominated hype.
She also ends with glimmers of hope, such as how AI can be used to preserve indigenous languages like those of Maori people, that would otherwise have disappeared (something she’d already covered for the MIT Review in 2022). This, by the way, is also the case with wasikan kisewatisiwin, a Canada-based AI tool which corrects unconscious bias or racism directed to Indigenous Peoples in written materials. True AI for Good, for once!
Careless People: “Careless” being the understatement of the year
Meta is currently on trial for not looking after its youngest users. It claims it could be fined $1.4 trillion, close to its market capitalisation. Whilst the real sum at stake could be more like $200bn, it has already been fined nearly one billion dollars recently, after a New Mexico court found it guilty of the same kind of complacency.
Going behind the scenes with Sarah Wynn-Williams, who had a long public affairs career at Meta before deciding to become a whistleblower, is shocking in that it confirms years and years (and years!) of despicable behaviour by the company’s leaders, thereby making even more predictable what’s happening to Meta now (although the mind boggles as to how they’ve got away with it for so long).
Even more shocking – as in a total surprise – is her decision to open the book with the tale of her shark attack as a child, in her native New Zealand. Her midriff was literally torn apart, then stitched back together by the local doctor. When she complained afterwards that she was still bleeding and in pain, her parents told her to stop being ‘soft’. She had to insist to be evacuated by helicopter and finally get the operation that saved her life.
So by the time she arrives at Menlo Park, Wynn-Williams is one tough cookie. But Meta’s shenanigans tested even her steel resolve, on a recurrent basis. Some key quotes:
On US/UK politics and Cambridge Analytica
“I was on a private jet with Mark [Zuckerberg] the day he finally understood that Facebook probably did put Trump in the White House [for his first term. Trump’s teams, led by Brad Parscale and leveraging Facebook user data illegally obtained by Cambridge Analytica, had bombarded facebook with tarteged anti-Hillary Clinton campaigns]”
“[Meta/Facebook’s then-number 2 Sheryl Sandberg’s] immediate response was not horror but that it was brilliant and innovative and do you think we might have a shot at hiring Trump’s guy Brad Parscale to come and work at Facebook?“
“Brexit has already happened and the leadership at Facebook are steadfastly avoiding acknowledging any role Facebook may have played in that” [Cambridge Analytica’s stolen data was also pivotal here]
“When the team posted that [republican politician] Ted Cruz’s Facebook strategy has “put money in all of our pockets and that’s what we call sweet, sweet coin”, underneath it read “Joel Kaplan and 22 others like this post”.” [Kaplan, a staunch republican Zuckerberg had hired to counter claims that Facebook was too ‘woke’ masterminded the rollout of political advertising on Meta platforms. Wynne-Williams would later accuse him of sexual harassment]
“If a local government notices false advertising on Facebook and asks to take it down, unless they have a law giving them power or prohibiting misleading ads that specifically applies to social media, they won’t act.”
On Facebook’s (failed) push into China
“Mark should not directly admit that Facebook wrote the censorship software in collaboration with the Chinese Communist Party… Facebook is operating illegally in China.”
On Facebook’s negligence facilitating genocide in Myanmar
“Community operations had written an internal report noting that 45 of the 100 most active hate speech accounts in South East Asia were in Myanmar” [and yet nothing changed. Facebook had just 2 moderators for that language. Ergo it] “enabled posts that led to horrific sexual violence and genocide. A lethal carelessness.”
On how teen depression was sold to advertisers
“Facebook is offering advertisers the opportunity to target 13-17 year olds across its platforms, including Instagram, during moments of psychological vulnerability when they feel “worthless”, “insecure”, stressed, defeated, anxious, stupid, useless and “like a failure”.” [all terms in quote marks as they’re keywords against which ads can be bought].
And so we come back to the present day’s trials, which feature Wynne-Williams’ fellow whistleblowers, like Arturo Béjar, or (hopefully) Frances Haugen, bith of whom have been key in revealing how little Meta cares about its young users’ well-being. As Wynne-Williams puts it in her book, perhaps the most surprising part of all that’s happened is this:
“The more they see of the consequences of their actions, the less of a fuck Mark and Facebook’s leadership give.”
At least the extent to which Meta is going to try to muzzle Wynne-Williams – including actively stalking her – shows just how right she must be. And fortunately, she has sued them back. Good luck to her…
Enshittification: How great, free things become shit, soul-sucking things
Cory Doctorow is one of my tech heroes. He’s just as good at analysing big tech’s abusive practices – with a special penchant for monopolies and anti-trust – as he is inventing science fiction scenarios around positive post-apocalypses or crypto-fuelled whodunnits.
Enshittification: Why Everything Suddenly Got Worse and What to Do About It falls firmly into the former category, of course. It begins by outlining the all-too-predictable stages of enshittification, namely:
Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die. I call this enshittification, and it is a seemingly inevitable consequence arising from the combination of the ease of changing how a platform allocates value, combined with the nature of a “two-sided market”, where a platform sits between buyers and sellers, hold each hostage to the other, raking off an ever-larger share of the value that passes between them.
The above is how he initially put it on his blog, in the context of TikTok, before expanding on the above three stages in the book. Some examples of enshittification he cites include:
- Amazon sold goods at a loss, which sucked in lots of consumers, locking them to the platform with ebooks and audiobooks that would only work in an Amazon environment; then with Amazon Prime; then by pulling in retailers who in turn sell at a loss on Amazon’s marketplace, and pay ads to make their products come out at the top of searches. Even if they’re shit…
- Facebook (them again!) was at first a great place to catch up with friends. Then once everyone was there, you couldn’t leave. Then with advertising, companies became dependent on Facebook. “It gave those advertisers a great deal, charging a pittance to target their ads based on the dossiers of nonconsensually harvested personal data they’d stolen from you.” Or how a pleasant user experience evolves into a dire one
- TikTok uses a technique they can “heating”, using “videos that Tiktok has inserted in order to make creators think that Tiktok is a great place to reach an audience.” This is how creators and brands alike are hooked in to the platform. But once they’re hooked, says Doctorow, the “heating” stops, and publishers have to work out by themselves how to get more attention. Guess how? By spending more money…
OK, so we’re doomed to enshittification. But how do we escape it?
Here, Doctorow somewhat predictably leans on some of his favourite core concepts. Respecting the end-to-end principle, whereby users get only what they asked for (think chronological social media feeds, as opposed to algorithmically-filtered ones); interoperability, or the possibility to take your data and content wherever you like, irrespective of platform (whereas alas we are inevitably locked in to the various tech platforms we use; and, above all, his absolute darling; antitrust laws. For Doctorow, breaking up monopolies is the number one way to stop enshittification. We like the idea; but who on earth light succeed in breaking up big tech in 2026? Google recently lost a US court case accusing it of being a monopoly… but didn’t have to change anything about its business, like potentially selling its Chrome browser, as a result.
But keep on trying, we must! So thank you, Cory, for the motivation ✊🏻
Les Assoiffeurs: They are taking all our water
Fabien Benoit and Nicolas Celnik’s book is not about tech per se; but it illustrates in such detail the degree to which the water sector too is obsessed with profit and “technosolutionism” – inevitably at great cost to the planet – that parallels are inevitable and useful.
There is a chapter about the tech sector – including a visit to a Scaleway data centre, courtesy of yours truly back in the day – but it’s not the most shocking, at least for those familiar with big tech excesses. We learn, for example, that Amazon (AWS), which already owns three data centres in Spain’s parched Aragon region (75% of which is becoming a desert), asked local authorities late 2024 if it could increase its water consumption by 47%, because of the “consequences of global warming”! Which it, of course, has no hand in exacerbating 🤔
Benoit and Celnik also astutely look at the chip-making sector, also a huge water consumer. STMicroelectronics, for example – European small fry compared with Taiwan’s monopolistic TSMC – is currently expanding its plant in Crolles, near France’s southern Alps. It will as such consume 33 million litres of water per day, more than three times its current rate; and 10 million litres more than the consumption of nearby Grenoble, one of France’s biggest cities.
Les Assoiffeurs – which literally means “the thirst-makers” – was inspired by the story of France’s mega-basins, the massive temporary lakes made to water industrial-scale farmers’ corn fields. Protesters denounce “a system of food over-production which benefits only a minority of farmers hell-bent on expansionist agriculture, made possible only by the monopolisation of water.” Indeed, the corn watered by these mega-basins mainly feeds cows in large-scale farms; and red meat is one of today’s biggest sources of greenhouse gases. Furthermore, such practices are regularly validated by Emmanuel Macron’s government. In agriculture as in tech, “bigger is better” often wins the day… even when it’s not better for the planet, or for society.
Perhaps the most shocking chapter, however, is the one about bottled water. We learn for example that in 2024, food and water giant Nestlé was found by French authorities to be illegally taking 19 billion litres of naturally-sourced water for 27 years, for its bottled water business. That same year, after the French Economy Ministry evaluated Nestlé’s fraud at €3 billion, it got away with a fine of €2 million.
Not to mention the bottled water sector’s plastic pollution – 4000-9000 billion litres of water per year, or up to twice France’s total drinking water consumption, are required to make the plastic bottles the water goes in – or the fact that 93% of water in bottles contains microplastics.
Above all, as with tech/AI, the book asks the question of how necessary all this is. It would cost $114bn to ensure universal access to clean tap water, according to a 2023 UN study; whereas the global bottled water industry was estimated at $270bn in 2021. Ergo this industry is not only unnecessary, it’s hugely polluting and wasteful; and yet, as the book points out, in times of drought, bottled water is often heralded as a saviour.
Similar shocks await in the chapter about water majors’ obession with technosolutionism. Companies like Veolia or Suez prefer to invest billions in technology like water purification, whereas investing in prevention – like encouraging farmers to use less pesticides, so less purification is needed – can be 65-87 times more cost-effective, according to a 2011 French local water authority study.
In water as in tech, whatever gets investors salivating – and investing – will, it seems, always get preference over what’s good for the planet. Thankfully, Les Assoiffeurs outlines some exceptions to this rule, such as Paris’ publicly-managed and not-for-profit drinking water system, which has become a global golden standard.
