Earlier this year, I read How Not To Use AI: 50 Contrarian Principles for the Imagination Age by Abi Awomosu. If you saw me during that time, there’s a good chance I fished it out of my bag to recommend it to you. Here are my notes.
It is a real counterpoint to the current AI narrative. We are currently being told a story about AI that is driven mostly by the big tech companies, but this book reminds us that there is always more than one story.
I found this book incredibly refreshing and I engaged with it in a way I’ve rarely done since my philosophy degree, including making notes in the margins when a point struck home or I had a counter-argument. I very much didn’t agree with everything she said, but it was very thought-provoking.
For that reason alone, I have been recommending everyone pick it up, to get a different take. Be warned though: it is a challenging read, in content and writing style, as well as being a daunting size.
Here are some of the points that resonated with me.
It’s very easy and often tempting to outsource your thinking to AI, and if you’re not careful, you could lose those skills.
She addresses this in her first law: “Don’t start with silicon—start with soul”. (Yes, that is an m-dash. Yes, you might find yourself wondering what support she had with writing this…)
She identifies this as “capitalism following its oldest pattern: find what people need, create dependency, sell it back to them”. Here, we are outsourcing thinking itself; and what makes it effective is that it feels like efficiency, like progress. But, she says, “when you start with AI you optimise for what it can deliver and abandon the intelligence you have that makes you irreplaceable.”
She has practical suggestions: be clear about what you are asking for before you start, spend time writing on paper before turning to AI, use AI only after you’ve developed complete thoughts. “Develop your cognitive capacity before adding AI amplification”.
She suggests that the term “generative” AI is misleading; it gives the impression that this generates something from nothing. In fact GenAI is trained on billions of pieces of content and reflects patterns back, as well as recombining information to create new patterns. She leans into the concept of AI listening to all this content, calling LLMs Large Listening Mediums (rather than Large Language Models): “Every Reddit frustration at 3am. Every Discord workaround revealing unmet needs… Every Amazon review describing exact pain points”.
Her point is that if we treat it as if it is intelligently generating genuinely new content, we get disappointing output, but when we lean into the way it has been trained, we can achieve better outcomes. For example, instead of saying “Give me 10 good marketing strategies” instead we can ask “What patterns do you see in why customers need but do not buy my product?” or “What patterns do you recognise in how our customers express their needs?” Instead of saying “How do we fix our broken process”, ask “what’s trying to emerge that our current process is blocking?”
This changed the way I ask questions of GenAI and it generates much more useful answers.
Later in the book she suggests doing “prompt poetry”. For example: “Show me how desire moves through digital spaces like wind through leaves” and later, showing it other worlds: “Where in nature does coordination happen without management?” One suggestion she has is ask GenAI to identify domains where your constraints don’t exist, or what pattern would make them irrelevant.
She also talks about asking for solutions from outside of your domain using the “pattern trinity”: one from nature, one from art and one from an unexpected source, e.g. children’s play.
She argues that these pattern-based questions strengthen your own pattern recognition capability.
Another thing she points out is that the nature of these tools mean they can work really well for different communication styles.
For example, she has ADHD and suggests some with ADHD may feel frustrated by trying to force questions into linear order – “learning to prompt properly” – and instead, it’s better to lean into stream of consciousness prompting, including all the tangential thinking. “By editing out your cognitive signature, you literally removed the competitive advantage—the unexpected connections that drive real innovation”. She has similar suggestions for Autism and Dyslexia.
This also ties in with a recommendation she makes earlier in the book to try speaking to AI rather than typing. She says research suggests that speaking activates broader areas of the brain; though like all research she mentions in the book no specific research is cited, and it’s not clear that any research into this does genuinely imply that speaking leads to a different quality of thought.
However, whatever the science behind it, I have definitely found that saying everything in an unstructured way can often be a very useful way to progress my thinking with AI.
One theme that comes up throughout the book is AI as extraction thinking. “The system extracts twice: first from unpaid creators whose work trains the models, then from underpaid workers who make the output acceptable.” She also says when you approach it as an extraction machine you get spiritually dead answers.
Her response is to think for ourselves, as a form of rebellion: “Every time you choose to think before you prompt you’re committing an act of resistance… they want your thinking externalised, captured, optimised and monetised. They want you dependent on tools they control, convinced you can’t function without them”.
However, the book is not internally consistent on this; later she talks about how to build yourself an advisory council using custom GPTs and says, approvingly, “after a month, working without your council will feel like thinking with half your brain”.
One the one hand, is that actually what we really want?! On the other hand, wanting to argue against some of her theses is one of the things that kept this book interesting for me, and helped me use my brain while reading it.
It ends with a proposal to grow an alternative garden. “Grow a garden so compelling that people naturally migrate.” I love this idea of how to fight back. What she explicitly means is a variety of things, from leaving companies that are surveillance-focused, to building digital platform cooperatives like Stocksy United and Resonate, to sovereign AI initiatives or distributed AI.
In a discussion with some peers about about this book we had a really interesting debate about whether we are moving back to the ’90s when everyone had a server in their office and we lose the value of centralised information, or whether that would actually be better.
In summary, she urges us to let AI accelerate our thinking and then take that learning into the world to build something real.
There is so much more in this book that I haven’t touched on; for example she talks about the computational inefficiency of the frontier labs’ models as being an architectural choice, and hallucinations as a reflection of how we are, and an opportunity.
She goes into the value of questions, “the creative act isn’t in judging outputs but in composing questions”, pointing out that the quality of your question determines the quality of your answers (true with or without AI), quoting the Sufi poet Rumi “Sell your cleverness and buy bewilderment”.
I really appreciated that her examples were very frequently women or people of colour; for example when she’s talking about visionaries who shaped our world, the list is Ada Lovelace, Octavia Butler and Grace Hopper.
This book is self-published, which I can understand: the pace of change in AI means you wouldn’t want to wait for a traditional publishing cycle to get this out there.
However, it could really have benefited from a good editor. It’s over 500 pages long and does not need to be; her arguments are strong, thought-provoking and refreshing but could have landed more strongly in a third of the length.
I did finish it, though if you are struggling to get through it, I’ll share that I did not get on with the third section of the book and she completely lost me around chapters 30-35, but I rejoined afterwards and continued to get value from it.
A few months after reading this, I am still thinking about it and applying some of the lessons. It definitely changed the way I think about and use AI, for the better.
The pattern throughout the book is reminding you of your own intelligence and what that brings. Done well, you are getting AI to amplify, sharpen, enhance, challenge, develop your own thinking. Not do the thinking for you.
It is a challenging read, both in a good way and in a not so good way. At times it is frustrating and the style can become grating. Some of the people I’ve recommended it to have not been happy with me afterwards (sorry Stephen!) but I still strongly recommend it.
Even dipping in and out will give value, and if nothing else, it will help keep your thinking nimble. It will remind you of the importance of continuing to think for yourself.
If you’d like to be notified when I publish a new post, and possibly receive occasional announcements, sign up to my mailing list: