For many, technology offers hope for the future―that promise of shared human flourishing and liberation that always seems to elude our species. Artificial intelligence (AI) technologies spark this hope in a particular way. They promise a future in which human limits and frailties are finally overcome―not by us, but by our machines.
Yet rather than open new futures, today's powerful AI technologies reproduce the past. Forged from oceans of our data into immensely powerful but flawed mirrors, they reflect the same errors, biases, and failures of wisdom that we strive to escape. Our new digital mirrors point backward . They show only where the data say that we have already been, never where we might venture together for the first time.
To meet today's grave challenges to our species and our planet, we will need something new from AI, and from ourselves.
Shannon Vallor makes a wide-ranging, prophetic, and philosophical case for what AI could a way to reclaim our human potential for moral and intellectual growth, rather than lose ourselves in mirrors of the past. Rejecting prophecies of doom, she encourages us to pursue technology that helps us recover our sense of the possible, and with it the confidence and courage to repair a broken world. Vallor calls us to rethink what AI is and can be, and what we want to be with it.
Some titles tell you nothing about the book itself - but The AI Mirror puts Shannon Vallor's central argument front and centre: that artificial intelligence, particularly generative AI such as ChatGPT, is not intelligence at all, but rather holds a mirror up to our own intelligence. As Vallor points out, your reflection in a mirror certainly looks and acts like you - but it is not a person.
This is a metaphor that works impressively well. It reflects (get it?) the total lack of understanding in systems that are simply reflecting back data from a vast amount of human output. That's not to say that they have no value, but we always have to be aware of their nature and their abilities both to produce errors as a result and to reflect our in-built biases, which we may consciously suppress but nonetheless come through in the data. To quote Vallor, these systems 'aren't designed to be accurate, they are designed to sound accurate'.
What Vallor tells us we have that AI doesn't is 'practical wisdom' or prudence - you might doubt this if you listen to some politicians (say), but the point is that we are able to engage this kind of filter where the AI lacks the ability - and though there can be tinkering at the margins when AIs get things badly wrong, it won't stop them continuing to trip up.
As someone with a science background, I usually find reading philosophy books a real struggle, as they are rarely anything but clear - however Vallor puts forward her arguments in what is usually well-worded, comprehensible English. The only exception is a near-obsessive love of the painful word 'valorize' (I don't know if this is nominative determinism).
One very small moan - Vallor makes use of science fiction parallels and a couple of times refers to a SF source called iRobot - I think this is meant to be Isaac Asimov's I, Robot, not the vacuum cleaner manufacturer.
There is some powerful stuff here, though at one point Vallor notes how we're all becoming poor armchair non-experts in subjects like climate change, a subject she refers to consistently throughout the book despite it not being her subject. But there is one big issue: for me this is the classic 'article stretched to be a book'. The key points are excellent and thought-provoking, but they are all made at length and could have been condensed into far fewer words with more impact. I am, nonetheless giving the book four stars for its readability and central argument.
The AI Mirror challenges us to reconsider the current trajectory of AI and its role in our society. Vallor argues that AI machines, by their very nature, look backward, limiting our ability to confront new and unprecedented challenges. She emphasizes the importance of phronesis—Aristotle’s concept of practical wisdom—as crucial for ethical decision-making. This wisdom, born from experience and our inherently messy, tragic nature, enables us to navigate the complexities of life. By relying too heavily on AI, we risk losing the very qualities that make us human. I refer to this process as 'The Great Flattening,' and Vallor similarly warns that the spaces where we can engage in moral deliberation and action are rapidly shrinking.
Vallor urges us to recognize that science and technology are not neutral instruments but are inherently normative, carrying moral and ethical implications. The current implementation of AI often overlooks these implications, treating humans as mere subjects to the whims of these technological artifacts. She challenges us to muster the courage—a virtue in itself—to raise our voices and demand that AI be aligned with our human needs, rather than allowing it to dictate our futures in ways that undermine our capacity for self-creation and moral judgment.
I recently re-read Plato’s Republic, and while reading it, I posted a few photos of some sections to Twitter, saying something to the effect that ‘Plato is insane.’ Vallor touches on similar concerns. In her concluding chapter, she writes:
“I don't believe in the devil, but if I did, I'd say the greatest trick the devil ever pulled wasn't to convince the world he didn't exist. I'd say it was to convince the world that things can't be truly humane and that beautiful ideas can't truly be materialized. Maybe I do believe in the devil. Perhaps his name was Plato.”
If you want a deeper reason why she is making this argument, read the book as she makes it.
Anyone concerned about the overwhelming deluge of AI-driven narratives that promote toxic positivity about the technology's role in our lives should read this book. Vallor provides the language and framing necessary to push back against the techno-utopianism, reminding us that while AI has its place in our future, it should never overshadow the beautiful tragic human condition.
I attended a great talk the other day by Dr. Shannon Vallor, exploring a more balanced perspective on AI.
She explored insightful research on where AI really is in terms of maturity. A targeted application seems to be where AI does work well, i.e., detecting/predicting breast cancer.
However, it is mostly rather unintelligent, which she demonstrated through rather lousy problem solving, i.e., the wrong solutions to simple problems such as crossing the river with a boat and a fox!
Given Vallor is a professor in the Department of Philosophy at University of Edinburgh, I really enjoyed the big picture questions and observations, touching on the idea that AI is a reflection of ourselves and echos back to us our own past biases and narratives (just like Narcissus).
It is an interesting and grounding problem. Are we falling for the hype and being a bit narssasitic about our collective creation and, more importantly, misunderstanding its potential and application?
Throughout the book Shannon highlights repeatedly and through various metaphors and explanations how "AI holds us frozen in place, fascinated by endless permutations of a reflected past that only the magic of marketing can disguise as the future." She speaks on how the tools needed for the immeadiate future cannot or should not be created from our past. She tears down ideas of superintelligence in emerging LLMs and reveals how they use datasets to predict the most probable response for the prompted input. Highlighting that the outputs of the LLM can only be forged from the data it initially digested.
An easy read and a good introduction to these thoughts / through relatable metaphors and use of accesible imagery Shannon paints this picture of the current state of AI.
Definitely not my favourite but it was an enjoyable read.
I couldn't finish this book. It rambled and repeated itself. The arguments were weak. The author came across as a luddite and technology hater. One of the worse philosophy books I've ever read.