blog: Book Review - I am not a Robot

src/content/blog/2026-06-19-i-am-not-a-robot.md

@@ -0,0 +1,60 @@
+---
+title: "Book Review: I am not a Robot"
+date: 2026-06-19T08:19:58.838Z
+slug: 2026-06-19-i-am-not-a-robot
+author: Thomas Wilson-Cook
+tags:
+  - book-review
+  - ai
+
+---
+I picked up journalist Joanna Stern's _I am not a Robot_ after hearing her on a podcast.  I like the public persona of Joanna, and was middling on her book's concept (use AI for everything for a year).
+
+I don't think I'm the target reader for this book.  Stern largely takes the claims of tech companies at face value when they tell us how AI will tell us the future of everything.  It read more like a "who's who at the zoo" than it did "but are they telling the truth though?"
+
+She opens the book most solidly on the use of AI technologies in medicine.  She shows in her own lived experience of a breast cancer screening exactly how machine learning and neural network tools are affecting the day-to-day experience of professionals and patients.  This is by far the strongest example in the book for me.  It's clear how a single-purpose, ruthlessly efficient bit of maths can work in a way that a multi-purpose, often tired human healthcare professional often cannot.  There are open questions about how we trust (and when we question) these tools, and how they can integrate into such critical, skilled settings.  And then how that might change in less critical healthcare settings (e.g. cancer screening vs. dentistry).  This conversation felt focused, and the end goal was clear: better detection, and earlier preventative care.  It's hard to argue against this.
+
+It's just sort of, everything after this example that didn't quite hit it for me.  We move rapidly onto consumer-facing artificial intelligence products and the boom in Large Language Models (LLMS) since the early 2020s.  This is less about empowering a specific profession with a specific tool, and rather more about promising a revolution to everyone while raising tens of billions of dollars.
+
+In this territory, my scepticism sits somewhere between "healthy" ad "prejudice".  I haven't seen the evidence that putting AI into everything makes everything (or even most things, or really even *some* things) better in the way that companies would like them to be.  If it was, it'd be an optional add-on, for which you could pay extra.  Like the beloved calculator analogy that gets trotted out[^calc], there aren't lobbies of people convincing your job to spend thousands of dollars on newer, better calculators.  Washing machines, also - you're welcome to do your washing by hand, and just as welcome to buy (or rent) a machine to do it for you.  Often generative AI is a mandatory feature ("feature"?) offered at the same price, or non-optionally linked to a price increase.  You know, standard enshitification stuff ([wikipedia](https://en.wikipedia.org/wiki/Enshittification)).
+
+
+It's here that I find Stern's presentation of "here's how it went!" to be too credulous.  During the book she interviews big figures (Bill Gates, Sam Altman), as well as various executives, and for the large part she reports on what was said in the interviews.  Rarely do we get evaluation or engagement with the ideas we're being told, or the story of the future we're being told.  
+
+Bill Gates tells Stern that LLMs will revolutionise healthcare.  He states that  every meeting with a primary healthcare provider is going to be four-way (two humans, each with their own LLM) - what does that mean?  Where does the data live?  What about low/no-internet environments?  Who checks that both LLMs came away with the same understanding?  Is the LLM data evidence one could use in court following malpractice or negligence?  Who pays for the LLMs?  What happens if I change the provider of my LLM, can I transfer data across?  Can the LLM interrupt the meeting to ask a clarifying question?  What if my new doctor's surgery has a different LLM provider to my previous one?  How do I request a copy of what my doctor's LLM data knows about me?  If I'm going to take an LLM into every doctors meeting for decades, these are questions that need to be answered.  
+
+The Altman interview is the most unhelpful example.  It comes near the end of the book, and we get two statements from him: 1. that AI is great, 2. that in the future AI will be even *more* great.
+
+Altman has a [well documented history of lying](https://www.newyorker.com/magazine/2026/04/13/sam-altman-may-control-our-future-can-he-be-trusted).  You can call it "telling different people what they want to hear", or you can call it "misrepresenting facts" - but if I tell my boss I will definitely spend the next two hours getting some work done, and then I tell my spouse that I can definitely go for a two hour lunch with them, I am both lying *and* telling different people what they want to hear[^1].  
+
+This is what I found most lacking in the reporting, because I think it's something missing from the wider ecosystem.  It's how billionaire tech executives are getting away with saying truly _wild_ things.  Like it'll double human lifespans in a decade (January 2035... start your clocks, [gizmodo](https://gizmodo.com/anthropic-ceo-hilariously-claims-ai-will-double-human-lifespans-within-a-decade-2000554601)), or we can expect Open AI to spend "trillions of dollar" on data centres ([The Verge](https://www.theverge.com/ai-artificial-intelligence/759965/sam-altman-openai-ai-bubble-interview)) - from a company that took $13bn in revenue and still lost $38.5bn ([Ed Zitron](https://www.wheresyoured.at/exclusive-openai-financials/)).
+
+If you come into the book sceptical, as I did, I found scant evidence that I was overlooking swathes of mounting evidence or experience.  I did not see many coherent visions of the future, there's a lot of betting on 'people will like our thing and get used to it', which may well be the case.  There is some research from the UK ([Alan Turing Institute](https://attitudestoai.uk/findings-2025/key-findings)), and much more research in the US ([Pew Research](https://www.pewresearch.org/short-reads/2026/03/12/key-findings-about-how-americans-view-artificial-intelligence/)) showing that people aren't, like, in love with unregulated AI and data centres.  In the US this has escalated to personal violence against Open AI CEO Sam Altman ([The Guardian](https://www.theguardian.com/technology/2026/apr/18/sam-altman-house-attack-ai)).
+
+If I came into the book as an AI booster I'd find nothing to make me evaluate my priors, as the worst people on the internet say.
+
+There is evidence that the people have opinions about the AI future we're being told is both inevitable and transformational.  As a species we've a remarkable capacity for cognitive dissonance, to hold two seemingly contradictory statements at the same time. But also we're able to just do things that are bad for us so long as they feel good enough, or the bad thing happens far enough away in time or space.  See smoking, eating too much red meat, domestic aviation, the use of child labour in garment productions.
+
+It's okay to not join the holy way of AI doomers vs. boosters.  It's definitely the more peace-keeping position to take, and Stern's writing style *is* of good quality - it deserves the chance to find its audience, and I genuinely think that audience sits much more on the AI-curious side of a spectrum, rather than the terminally-online opposite.  
+
+I think it'd be a frankly out-of-scope expectation to have Stern solve the contradictions at the nature of the human condition.  That's much more of a Pope kind of job, and he _is_ giving it a go ([Wikipedia](https://en.wikipedia.org/wiki/Magnifica_humanitas)).  
+
+We also just flat out _don't know_.  I can ask for more evidence, and of higher quality, but right now (June 2026) I think I'm asking for something that almost literally nobody has.  Given how new generative AI and LLMs in general are.  We simply haven't had time to do the research yet, and it's felt like we're waiting for the AI investment bubble to pop.  
+
+So while I'd love answers (really, I would), I'm content to just ask that these things are considered seriously in the same conversation as "I tried a robot masseuse and it was actually okay".
+
+But in a couple of occasions, namely around robots, Stern presents researchers' opinions as either caveats or equal to that of company executives.  Tesla's claim that its human robots will be worth seven squillion dollars or whatever (and also please buy our stock, and also please don't look at our [CEO's track record with predicting autonomous driving](https://en.wikipedia.org/wiki/List_of_predictions_for_autonomous_Tesla_vehicles_by_Elon_Musk)) is in the book.  These extraordinary claims are followed by the caveat that experienced robotics researchers say that we're likely decades away from what's being described.  These two statements can't co-exist, neither can be true at the same time.  Tech CEOs have incentives to lie, and researchers probably have incentives to be conservative in their public statements (literally one profession should have this).
+
+Stern moves on from points like this a bit too quickly for me.  I think that the ability to evaluate confidently stated, practically outrageous predictions about the future *is* a core part of modern AI literacy.  These contribute hugely to the anxiety around AI (see above), and another journalist plainly stating or reporting on them isn't helping.  I think the lack of critical engagement, even if it's to explain the magnitude of the prediction and to soften the confidence, continues to create an environment where people really can just sort of say whatever.
+
+It's a shame that Stern has walked into a situation where the temperature has already been cranked up this high.  It *shouldn't* be the job of journalists to deescalate the situation, and to undo the damage done by the world's wealthiest losers[^2].  But that *is* the situation she's in, and immersed herself in for a year, and then wrote a book about.
+
+Look, read it, Stern's a writer with a strong voice, but don't expect it to change your mind about literally anything.
+
+---
+
+[^1]: At the time of writing, lunching is not my job.  Don't mix work and passion, that's what I say.
+
+[^2]: Actually if you make your millions from petrochemicals you're also a loser, but at least you know deep down you're doing moral gymnastics to justify forcing people to migrate across the globe and experience famine, as well as destroying entire ecosystems.  All of your oil is *definitely* going to hospitals and fertiliser.  Or you feel no empathy or connection to humanity and the money's good, i.e. you are a sociopath.  Actually maybe these are the same type of losers.
+
+[^calc]: A common argument around AIs is that nay-sayers are under-representing historical transformations, and that the current wave of generative AI is comparable to previous technological transformations.  The argument goes that just like how we don't have to do maths by hand anymore, we use calculators, won't have to do so much reading/writing as before using new LLM technologies.