Nell’occhio dell’algoritmo: Storia e critica dell'intelligenza artificiale

by Matteo Pasquinelli

Che cos’è l’intelligenza artificiale? In molti credono si tratti di un progetto per replicare l’intelligenza umana basato sulla logica segreta della mente o sulla fisiologia profonda del cervello e delle sue reti neurali. In questo volume, al contrario, si sostiene che il codice interno dell’IA non imita l’intelligenza biologica, ma l’intelligenza del lavoro e delle relazioni sociali, come tra l’altro confermato già dai Calculating Engines ideati da Charles Babbage a metà Ottocento e ancor oggi dagli algoritmi per il riconoscimento di immagini, per la sorveglianza e per la generazione di testi. L’idea che un giorno l’IA possa diventare autonoma o senziente è perciò pura fantasia, poiché gli algoritmi hanno sempre emulato nella loro struttura la forma delle relazioni sociali e l’organizzazione del lavoro, e il loro scopo rimane cieca automazione. Alla luce di ciò, l’autore invita scienziati, giornalisti e attivisti a sviluppare una nuova consapevolezza dell’IA, riconoscendo che il suo “mistero” non è altro che l’automazione del lavoro su grande scala, non l’intelligenza in sé.

Genres: Nonfiction, Artificial Intelligence, History, Technology, Philosophy, Science, Politics

268 pages, Paperback

Published May 1, 2025

Reviews

[Gavin Kierulf · Rating 5 out of 5]

This was a deeply interesting book, as close to a riveting page turner as an academic and information-rich book can get.

First and foremost, I found the treatment of the 'algorithm' and its long history incredibly interesting and forced me to re-visit some of my thoughts and feelings about math (a subject I've never been particularly good at). Additionally, the earlier portions of the 'social history' were of particular interest and I found them more engaging and easier to grasp with my foreknowledge of the time-period and language of labour. This is as opposed to the chapters on the development of 'AI' through the 1940-60s and its related disciplinary lexicons which I had to spend some more time with.

Instead of pointing to any particular flaws, by way of critique I’ll instead offer thoughts on what I thought were more overlooked aspects of ‘AI’ in The Eye of the Master.

The first somewhat overlooked (or at least not fully fleshed out) aspect that jumped out at me is how Pasquinelli doesn't fully develop a challenge to the adoption of 'intelligence' in ‘AI’ discourse. Of course, it is mentioned, and its genealogy is traced expertly, but knowing the ‘whys’ and ‘hows’ of the word being adopted is not the same as a critique. That being said, it doesn't escape me that a full critique of 'intelligence' itself was outside the scope of this book's endeavour – nor was it the point.

Regardless, the text that immediately came to mind to fill this gap of an ‘intelligence’ critique was Technically Man Dwells upon This Earth by Ulysse Carrière. As Carrière puts it, "intellect does not operate with a model, only the sensible does. There is something that does not require an input, and that is the intellect. As such, the intellect lies beyond the realm of techne" (Carrière 2023, p.19). For Carrière, this challenge to the ‘intelligence’ of AI brings up "the immanent necessity of understanding Thinking as a thinking of the Beautiful” (Carrière 2023, p. v). This is to say the appropriation of the language of (and in Pasquinelli’s genealogy, a genuine belief in) ‘intelligent’ ‘AI,’ immediately leads to questions over the creative impulse beyond the more ‘social’ incapability of ‘AI’ and instead towards the transcendental act of ‘creation.’ These notions stayed so heavy in my mind while I was consuming Pasquinelli’s book because of the importance of recognizing the potential ‘threat’ of ‘AI’ poses to distinctly creative modes of labour, so that in the face of this crisis(?) "art is being tasked with the creation of the unthinkable. A new art has been made necessary" (Carrière 2023, p. 37). The crisis of creation (as opposed to a ‘creative crisis’) which ‘intelligence’ in ‘AI’ poses, while (understandably) absent from The Eye of the Master, I think remains a pressing concern for critical ‘AI’ discourse.

The other, somewhat related, aspect that I felt could have been developed further by Pasquinelli is tying into the present moment more with today’s ‘AI’ systems. It is only in the conclusion of the book where Pasquinelli turns to these more theoretical concerns which are, admittedly, outside the broader scope of his ‘social history’ (rather than a social present). Of course, Pasquinelli’s book itself is a (social) history not a contemporary account, so this is much less a critique and more a personal theory-brained nitpick. However, where the conclusion does dip into this contemporary and urgent discourse, Pasquinelli deftly sums up not only a major theme of the book, but also the discourse more broadly declaring,"the first step of technopolitics is not technological but political" (Pasquinelli 2023, p. 253).

I cannot recommend this book enough, Pasquinelli writes about a very technical subject matter concisely, and conveys a complex history with all its attendant theories, debates, and complications in a satisfying a highly readable manner. I feel better prepared to engage with concepts of and around ‘AI’ having read the book, and I am excited to potentially follow his notes into further reading on the topic.

[Ferda Nihat Koksoy · Rating 4 out of 5]

"İnsanlar ses aralıkları veya nota sistemi hakkında en ufak fikir sahibi olmadan şarkı söylüyor veya müzik yapıyor. Sayı sayma da rakam mefhumundan eskidir.(T.Macho)
Buhar makinesinin icadı termodinamik disiplinini ortaya çıkarmıştır yoksa termodinamiğin gelişmesi buhar makinesinin icadına götürmemiştir".

"1823'te Babbage'ın modern bilgisayarların habercisi olarak nam salmış Fark Makinesi (Difference Engine) ticari bir saikten doğmuştu: Astronomide ve İngiltere'nin deniz ticaretindeki üstünlüğünün idamesinde elzem olan logaritma hesaplarını otomatikleştirip, hatasız logaritma tabloları satma arzusundan. Bilgisayımın makineleştirilmesini teşvik eden sorunların en önemlisi, açık denizde boylam hesabıydı. Küçük mekanik hesaplayıcılar zaten mevcuttu ama otomatik olmadıkları gibi yalnızca temel matematik işlemlerini çözebiliyorlardı. Babbage, karmaşık bir logaritma hesaplayıcısını buhar makineleriyle sürdürülecek kesintisiz harekete bağlamayı düşündü; böylece sadece bir hesap aygıtı değil, hesap işini endüstriyel ölçekte yerine getirebilecek bir hesap makinesi geliştirecekti - sınırsız verimlilik ve dizginsiz ekonomik büyüme fantezileriyle birlikte."

"Makine zekâsı tabiri en az dört anlam içerir: (1) insanın makineye dair bilgisi; (2) makinenin tasarımında cisimleşen bilgi; (3) makinenin otomatikleştirdiği insan işleri; (4) makinenin kullanımı sayesinde önü açılan, evrene dair yeni bilgi."

"Endüstriyel makine çok etkileyici bir mamuldür zira enerji ile madde, bilgi ile bilim, en önemlisi de sermaye ile emek arasındaki ilişkileri tek bir nesnede üst üste toplar. Bu anlamda endüstriyel makine kapitalizmin birçok çelişkisinin cisimleşmiş hali olmanın yanı sıra somut bir toplumsal ve ideolojik mücadele sahası gibidir. Teknolojinin siyasetteki merkezi yerine duyulan ilgi günümüze kadar devam etmiştir; akademik tekno-determinizm de, sermaye cephesindeki tekno-çözümcülük de teknolojiyi siyasetin merkezine alır.
Ancak bu kitapta açıklamaya çalıştığım üzere, emeğin işbirliği başta olmak üzere toplumsal ilişkiler teknik ve siyasi gelişmenin 'motoru'dur.
Yapay Zekâyı toplam-işçinin, genel zekânın, Marx'a göre endüstriyel üretimin başaktörü olan Gesamtarbeiter'ın (Genel zekâ; kol ve zihin toplamı) son görünümü olarak düşünmek mümkün."

"Sabit sermayenin (komplike makine) gelişimi, genel toplumsal bilginin ne ölçüde doğrudan bir üretim gücü haline geldiğini, dolayısıyla bizzat toplumsal yaşam süreci koşullarının genel zekânın güdümüne ne ölçüde girdiğini ve buna uygun olarak ne kadar dönüşüm geçirdiğini gösterir.
Bilgi ve beceri birikimi, yani toplumsal beynin genel üretici güçlerinin birikimi, böylelikle emeğin aksine sermayede massedilir ve bu yüzden, asıl üretim aracı olarak üretim sürecine dahil olmasından ötürü sermayenin -daha belirgin olarak da sabit sermaye- temel bir özelliği olarak belirir. Demek ki makine sabit sermayenin en uygun biçimi olarak, sabit sermaye de bilumum sermayenin en uygun biçimi olarak belirir." (Karl Marx, Grundrisse, 1853)

"Bizzat insan emeği tamamen mekanik hale gelmekte, çokyönlü bir belirliliğin parçası olmaktadır. Ama emek soyutlaştıkça, insanın kendisi de salt soyut etkinlik haline gelmektedir." (Hegel, 1820)

"20.yy'ın sonundan beri emeğin yönetimi toplumun tamamını 'dijital bir fabrika'ya çevirmiş; arama motorlarının, çevrimiçi haritaların, mesajlaşma uygulamalarının, sosyal ağların, esnek ekonomi platformlarının, esnek ulaşım hizmetlerinin ve nihayet tüm bu hizmetleri otomatikleştirmekte kullanılan YZ algoritmalarının yazılımlarıyla aynı biçimi almıştır. Bugün YZ'yi dijital toplumun merkezileşmesindeki ve toplumun her kesitinde işbölümünün denetlenip düzenlenmesindeki bir üst aşama olarak görmek mümkün."

"Yapay Zekâ projesi aslında hiçbir zaman tam anlamıyla biyomorfik olmamış (yani doğal zekâyı taklit etmeyi hedeflememiş), örtük biçimde sosyomorfik mahiyet göstermiştir (yani, kontrol edebilmek amacıyla, toplumsal işbirliği ve kolektif zekâ biçimlerini kodlamayı hedeflemiştir). Zekânın otomasyonunun akıbeti, özerklik yönündeki siyasi yönelimden ayrı düşünülemez: Nihayetinde YZ projesine biçim ve ivme kazandıran etken, toplumsal zihnin kendini örgütlemesi olmuştur."

"Yapay sinir ağlarının önünü açan düşüncenin temelinde uzun soluklu bir ihtilaf yatıyor: insan algılamasının analitik olarak temsil edilmeye, dolayısıyla makineleştirilmeye yatkın bir idrak edimi olup olmadığı tartışması. Bu soru etrafındaki tartışma 1940'larda sibernetikçiler ile Gestalt ekolü arasında alevlendi: Sibernetikçiler algı alanının tamamının basit elektrik röleleri gibi makineler tarafından hesaplanabilir olduğunu iddia ediyor, Geştalt ekolü ise bir makinenin insan zihninin karmaşık sentetik yetisini hiçbir zaman taklit edemeyeceğini savunuyordu.

"YZ büyük altyapı modelleri, son derece geniş bir kültürel miras, kolektif bilgi ve toplumsal veri havuzu üzerine inşa edilmiş olmaları bakımından, sanayi çağının "genel zekâ"yı makineleştirme hayaline en çok yaklaşan aygıtlar. Altyapı modellerindeki makine öğreniminin önemli özelliklerinden biri de, tekil görevlerin otomatikleştirilmesi, kültürel mirasın kodlanması ve toplumsal davranışların analizi arasında teknik açıdan hiçbir ayrım olmaması: Bunların hepsi aynı istatiksel modelleme süreciyle (otomasyonun otomasyonu) hayata geçirilebiliyor."

"2021 tespitlerine göre, derin öğrenmedeki hata düzeltme tekniklerinin bir hesaplama sınırına ulaştığı, enerji ve donanım kaynakları maliyetlerinde büyük şirketlerin bile bir süre sonra karşılamakta zorlanacağı muazzam artışlar olmadan bu teknikleri daha fazla geliştirmenin mümkün olmadığı ortaya konmuştur (Neil Thomson).
Hesaplama kısıtları göz önüne alınırsa, mevcut YZ'nın kötücül "süper zekâ"ya dönüşmesi gibi bir tehlike de söz konusu değil."

"Günümüzde otomasyon, basmakalıp hale gelen insansı robot imgesindeki gibi tek tek işçilerin emeğinin otomasyonuyla sınırlı değil, esnek ekonomi platformlarında olduğu gibi fabrikadaki patronların ve yöneticilerin işinin otomasyonunu da içeriyor. Lojistik devlerinden (Amazon, Alibaba, DHL, UPS vb.) taşıma ve teslimat alanındaki büyük şirketlere (Uber vb.) ve sosyal medyaya (Facebook, TikTok, Twitter) kadar birçok sektörde, platform kapitalizmi denen otomasyon biçimi aslında işçilerin yerini almıyor, bilakis işçi sayısını artırıyor ve onları farklı bir tarzda yönetiyor. Bu defa mesele emeğin otomasyonundan ziyade yönetimin otomasyonu. Bu yeni algoritmik yönetim koşullarında hepimiz, her çeşit küresel kullanıcıdan, makineyi çalıştıran görünmez işçilerden, bakıcılardan, sürücülerden ve kuryelerden oluşan dev bir otomatın bölünmüş işçileri muamelesi görüyoruz. YZ'nin bütün işlerin yerini alacağı endişeleri üzerine dönen tartışmalar yanlış bir varsayımdan yola çıkıyor: Platform ekonomisinde algoritmalar gerçekte yönetimin yerini alıyor ve güvencesiz iş sayısını artırıyor. Esnek ekonomideki kazançlar geleneksel yerel sektörlere kıyasla hâlâ çok düşük olsa da, bu platformlar dünya çapında aynı altyapıyı kullanarak tekel konumlarını sağlamlaştırdılar. Sonuç olarak, bu yeni "patron"un kudreti tek tek işlerin otomasyonuna değil, toplumsal işbölümünün yönetilmesine dayanıyor. Alan Turing' in öngördüğünün aksine, robotun yerini aldığı ilk taraf işçi değil patron (tekel olamayan)  oldu."

"Bizim asıl ihtiyacımız, tüm sorunları teknolojiyle çözmeye kalkmak veya tersine teknolojiden tümden vazgeçmek değil, topluluklara ve kolektife ihtimam gösteren, failliği ve zekâyı asla tümüyle otomasyona terk etmeyen bir tasarım, planlama ve icat kültürü yaratmak. Teknopolitikanın ilk adımı teknolojik değil politiktir. Ancak, YZ'yi aşacak bir siyasi epistemoloji projesi, tarih boyunca başvurulan soyut (matematiksel, mekanik, algoritmik ve istatiksel) düşünce biçimlerini dönüştürmek ve bunları doğrudan eleştirel düşüncenin alet kutusuna katmak zorunda. YZ'nin epistemolojisiyle ve bilgi hafriyatçılığı rejimiyle mücadele etmek için farklı bir teknik zihniyetin, kolektif bir "karşı-zekâ'nın öğrenilmesi gerekiyor."

"Çağımız dijital veri çağı ve bu çağda matematik siyasetin meclisi haline geldi. Toplumsal yasa modellerle, teoremlerle, algoritmalarla iç içe geçti. Dijital verilerle birlikte matematik, insanların teknolojiyle eşgüdüm sağladığı başat araç haline geldi. Matematik her şeye rağmen bir insan faaliyetidir. Her insan faaliyeti gibi o da hem özgürleşmenin hem baskının imkânlarını barındırır."
(Politically Mathematics kolektifi manifestosu, 2019)

[Jooseppi Räikkönen · Rating 3 out of 5]

So I found this book extremely profound and interesting. Further, I even most likely buy its core message. A couple of large complaints though stop me from fanboying it.

First, the book stops point blank before some of the biggest leaps in A.I. While this is understanable, especially for a purported "history", I was still troubled by the sense that some of the A.G.I. proposals combining active robotics together with LLMs etc. aren't sidetrackable with the Marxist social intelligence retort.

Second, I found it strange that there was no reference to Ilyenkov or Soviet Cybernetics. The author's perspective is the whole raison d'dêtre of the activity school. (though there was an extremely brief bit about Vygotsky)

Finally, it just seems to (and I know this is kind of a lazy criticism) not navigate the trappings of historical ontology well enough to shed the doubt that it really constitutes crude historicism about A.I.. It seems overly pre-determined to me to say that we couldn't discover for example the theory of the brain through an investigation of social relations with purely historical evidence. After all, that is essentially the story of thermodynamics and the steam engine.

[Malcolm · Rating 4 out of 5]

There are few issues in these times that provoke debate quite as much as AI, in some circles at least. In my world of academia it is seen as a profound challenge to academic integrity; elsewhere it is celebrated/feared as the end of mundane labour; for others it presents an existential threat (after all, we’ve seen Terminator). And yet I sat in a meeting recently with some of the doomsayers, as workmates from our medical technology development unit explained the essential role of AI in medical treatment, and from media production outlined how vital it is to make sure their students are well versed in these technologies. It is as if AI is, in part at least, the moral panic du jour. More generally though the discourse is torn between those whose technological fetishism holds out for the machines that will finally act as if they are human brains, and those who see the technology as more dangerous than promising.

There is, of course, an awful lot written about AI, its development, its potential, its dangers, and more, as technophiles and technophobes position themselves as the voices of expertise, yet one of the more interesting things about these ‘discussions’ is their comparative silence on what is meant by ‘intelligence’. It is almost as if there is an individualistic, anthropocentric vision that reduces ‘intelligence’ to the functioning of the human brain. This is where Marcello Pasquinelli comes in most productively, by making two convincing arguments: the first is that there is not much about algorithms or many other aspects of AI that is new (although there is when it comes to things like power and scale), and that there is a much more significant dynamic in explorations of work and labour that sees ‘intelligence’ as social, as shared, and as collective.

The case turns on these two aspects – that there is a long history of the stuff, the components, of AI, and that the seizing of social collective knowledge is the stuff of growing, centralising, capitalist economic power. So, much as this is labelled a social history, it’s not as social historians would understand that term, where we’d look at who participated or engaged with what practices, how, in what way, and to what end. It is more that this is an intellectual history of AI, weaving some surprising strands into the field and building an eclectic vision. The closest Pasquinelli gets to detail of the current debates, engagement, and panic about AI is a case he makes towards the end of the argument that in many cases algorithms and the systems they sustain can be turned to many ends, as data and other material is sorted, recast, and directed to specific goals.

Pasquinelli firmly rejects techno-centric analyses (so in that sense this is social history). As a result, the opening section lays out the fundamental aspect of the argument that technological tools are the product of social need (need being my word, not his, and I recognise it as remarkably problematic). This materialist view is wrapped in a history of aspects of mathematics (the fundamental tool here is a series of equations – algorithms – instructing a process), focusing on the shift from counting to calculation. While it may seem arcane, this notion of calculation as a consequence of social requirements is essential to his case that AI is a form of engagement with the social, with the collective knowledge.

The first part of the argument is where he delves most explicitly into this position – and I confess his grappling with the notion of the ‘general intellect’ that Marx laid out in some of his earlier work means I am in a conceptual and philosophical space which is quite comfortable. Eeven though by the time we get to Capital Marx has shifted this to a notion of the general worker, and tied it into notions of abstract labour, what he didn’t get away from, and the aspect that is vital for Pasquinelli’s argument, is the idea of the collectivization of workers’ knowledge into a system where that knowledge is alienated from them, systematised, and deployed to the benefit of those with power. This section takes early technological writing – notably Babbage and Ada Lovelace – and recasts it, shifting Babbage’s approach to one that stresses workers’ interests. There is important attention paid to the ‘machinery question’ and working class resistance to the technology of the factory. With that comes big questions about technology and the labour theory of value that add significant nuance and present aspects (the labour theory of technology) the implications of which need careful working through. Yet the core of this aspect of the analysis is precisely the question of the political and social struggles, and the very real point that these were social questions, not technological developments in the abstract. That is to say, this is a materialist philosophy at work.

The second aspect of the case shifts focus to a field more likely to be more usual for discussions of AI. Here he explores the developments from the mid-20th century looking at moves to forms of technological self-organisation. He relies heavily on developments in cybernetics, but even so there are surprises. The case draws on the expected worlds of developments and debates in engineering, but weaves in the growing power of psychometrics in psychology, Hayek’s work on the self-organization of markets, and the push to the automation of perception. This means Pasquinelli is able to explore both the technologies of self-organisation and the challenges of working with either an excess or absence of information. This is a more technological and less obviously philosophical discussion, although there are powerful epistemological debates that run through the advocacy and finessing of specific technologies, especially as these technologies are intended to work with how the world and knowledge (‘intelligence’) works.

Weaving these together allows Pasquinelli to explore the automation of the general intelligence, to make the compelling case that this current form of automation (here’s the machinery question again) seems to be a factor in the growing precarity of workers’ jobs, and a real threat to managers’ jobs. His point seems to be that the management of labour is more likely to be automatable than the conduct of work; even though tools such as the Mechanical Turk break jobs into discrete pieces the effect is precarity, not redundancy. It’s a fabulously counter-intuitive conclusion, but one that is worth a much closer look.

Pasquinelli’s historicised and materialist case for understanding AI means that cutting off his case around the turn of the 21st century does not undermine its power; to a large the developments of recent years have been those of quantity – speed and reach – rather than quality – tech that is doing fundamentally new things. It also means we need to (re)think responses and political struggles around AI, and the class politics of these technologies as something that has been missing from many of the debates.

Crucially, also, this is far from the last word on AI, on systems thinking and operation, and on the dynamics of self-organization – but it is a good productive and provocative step down an insightful path through those issues. It powerfulness and productivity in that sense is in large part a consequence of the range of approaches and insights Pasquinelli weaves together into a politically sharp, intellectual history of AI. This makes it a challenging read, but a valuable one.

[Kaleigh · Rating 5 out of 5]

THE book on AI. I keep trying to find a more updated version of a book like this (it’s only a couple of years old but the discourse is always changing), but haven’t found one yet. So if you’re going to read a book about AI make it this one. I promise!!!

[Jonna Higgins-Freese · Rating 1 out of 5]

I was disappointed, perhaps because saying that AI "is constituted by the intelligence of labour and social relations" or that "all labour is logic," makes me go, sure, okay - but what does that mean, when it's at home? Honestly, probably the best takeaway for me was a list of other books he references that might actually give enough concrete information to make his points comprehensible.

The title comes from a quote from Engles about their being no "eye of the master" of the London and North Western Railway, b/c it is too large, a critique that seems to me substantially undercut by LEAN methods like the Andon cord and also the vastly increased regimes of surveillance even in large operations (keystroke tracking, GPS speed tracking, Amazon warehouse body tracking, etc.) (4).

"This book is intended as an incursion into both the technical and social histories of AI, integrating these approaches into a sociotechnical history that may identify also the economic and political factors that influenced its inner logic. Rather than siding with a conventional social constructivism and going beyond the pioneering insights of social informatics, it tries to extend to the field of AI the method of historical epistemology (12). Individually, i know what each of these words and concepts mean, but I can't make head or tail of them as they're used in these sentences.

Some babbly stuff that could likely be interesting if I had any idea what he's actually talking about: that deep learning did not arise from theories of cognition but from pattern recognition, which had been opposed to theories of cognition in early AI. Okay, but so what? (14).

"The first artificial neural network - the perceptron - was born not as the automation of logical reasoning but of a statistical method originally used to measure intelligence in cognitive tasks and to organize social hierarchies accordingly." (21) - again, I kind of get where he's going here, but it's so vague and abstract as to be not actually comprehensible.

"In early nineteenth-century England [and late nineteenth-century United States, according to Dava Sobel, The Glass Universe -- and also, to talk about this at this point without mentioning _Hidden Figures_ is just weird], 'computer' was not the name of a machine but of a human - name an office clerk, often a woman, who had to make tedious calculations by hand for the government, the Astronomical Society, or the Navy. At times 'computers' were also working from home, receiving stacks of numbers to calculate and sending back results by mail: this was literally the first historical occurrence of a computing network that took the form of domestic labour and probably involved further family members. with the aim of streamlining this time-consuming and error-prone process, the polymath Charles Babbage had the idea of replacing the repetitive work of may 'computers' with an automated machine powered by steam" (52).

Babbage's Difference Engine "was born out of a business ambition - to automate the calculations of logarithms and sell error-free logarithmic tables, which were crucial in astronomy and for maintaining British hegemony in maritime trade" (52). "The idea of the automatic computer . . . emerged out of the project to mechanize the mental labour of clerks rather than the old alchemic dream of building thinking automata - although the latter narrative would often be used, . . . to masquerade the former business" (52).

IBM punch cards used in the census of Jews in Nazi Germany.

One genealogy of computation and AI follows artificial neural networks for pattern recognition, another to automate hand calculation (53).

"as an expression of the division of labour, computation watched over the unfolding of industrial capitalism from its very outset" (maybe the actual books on Babbage provide the concrete examples that support this claim). (54).

55 - one early desire to automate computation arose from the French Revolution's attempt to change the right angle to 100 degrees, necessitating recomputation of log tables (55).

"The extended list of knowledge models and modalities of knowledge production that crossed the industrial age should at least include: the type of knowledge that is represented by the act of invention of a machine; the division of labour that inspires its design; the know-how of engineering and the symbolic language that is necessary to describe the mechanism; hard sciences such as mechanics and thermodynamics; non-technical disciplines such as political economy; the metrology of manual and mental labour and the instruments to measure them; collective knowledge as embodied in both machinery and social relations (the so-called general intellect); educational movements such as the Mechanics' Institutes; political campaigns such as the March of Intellect; and, finally, popular mythologies around automata such as the Mechanical Turk." (83).

Current AI as "the automation of management" (250).

Questions how it would be possible to redesign large-scale monopolies of data and knowledge (251).

Some interventions: turkopticon.net, exposing.ai, politicallymath.in, https://www.ai.hps.cam.ac.uk/

[Maria Aleksandrova]

Interesting topic but not very accessible to a casual reader. Made me fall asleep not once but twice.

[Hugo · Rating 2 out of 5]

Have you ever been blue balled by a book? I can now say that I have.

Although the book gives us some really interesting insights on the history of AI, the political parallels that it makes along the way are put in your face so closely that I had to take a literal step back in order to not get hit by them.

Honestly if you want to write a political essay, go for it, it is supposed to be a social history of AI after all. And that's a shame because I do agree with the core message, but that message gets lost in all this discourse... And at the end of every chapter my honest reaction is "yeah okay, so what?". The author literally bombards you with so many facts, names, concepts and everything he can than by the end you're like "wait wth was this chapter about?". Great showing off in terms of knowledge and really interesting, but then what is this book really about?

The book wants to enlighten us on the true nature of AI and the sociological hierarchy it will inevitably create (again, I do agree with that), but giving us a rather uninspired rehashed version of The Kapital gets us nowhere. The book just shoots on the ambulance, writes a political note on the side and calls it a day.

Also, is it too much to ask to make sentences shorter than a full page? I shouldn't be feeling out of breath after reading ONE paragraph.

[draxtor · Rating 5 out of 5]

So good, so much history, so much context.
So important!
Lots of aspects of so-called AI under the microscope including the question of working more/less when everything can be automated in the warm bright future.

Unfortunately nobody in my circle of RL (or Second Life) friends who are enthusiastically telling me that AI will make us WORK LESS AND ENJOY MORE LIFE AND MORE CREATIVITY (and citing Ben Affleck telling them that MORE GOOD WILL HUNTING WILL BE PRODUCED BECAUSE LEVELING PLAYING FIELD blablabla) will read this book.

In case they read this "review" let me give them a little TL'DR in an algorithmic 4 steps:

1.) watch Charlie Chaplin's "Modern Times"
2.) ponder the question if automating the part that you do in a division-of-labor regime will result in the machine letting you work less or forcing you to work more
3.) consider reading the book
>>> if 3.) is YES I AM GONNA READ IT = exit this algo, shut down stuff, disconnect, organize and KEEP READING
>>> if 3.) is NAAAH I AM GONNA KEEP SCROLLING THROUGH MY SOCIALS BECAUSE I JUST GOT OFF WORK AND I AM EXHAUSTED AND I NEED TO CHILLAX FOR A BIT BEFORE REPRODUCING FOR THE NEXT DAY AT WORK = go back to 1.)
>>> if 3.) is MAYBE SOME DAY IN THE FUTURE = go to 4.)
4.) repeat from step 1.)

Ok what have we learned? You are correct: I FAIL at writing code!

[Francesca · Rating 4 out of 5]

very good , there was a sharp switch to purely scientific in the final 3 chapters which felt disengaging , much preferred the balance between theory, history and science in the first half of the book

[RoaringRatalouille · Rating 5 out of 5]

This is an excellent book and recommendable to all those interested in the history of what we have come to call "AI" today. It is recommendable because it advances a number of counter-intuitive claims about the history of AI that are nonetheless plausibly explained and argued for. What follows is a rather lengthy summary of what to me are the most essential and interesting arguments made.

First, Pasquinelli starts his book by reiterating a commonly heard statement encountered in AI textbooks, speeches by tech gurus, and public discourse more broadly: the idea that AI is the quest to "solve intelligence" that is to be located within the human brain. Put differently, the dominant view describes AI as an imitation of intelligence in a biological sense. The author, to the contrary, claims that AI emerges not as a result of imitating biological intelligence, but the intelligence of labour and social relations. To make this argument, Pasquinelli employs a historical epistemology that centers on the "dialectical unfolding of social praxis, instruments of albour, and scientific abstractions within a global economic dynamics". Very interestingly, in the book's first chapter, Pasquinelli shows and argues that algorithmic thinking - as opposed to mainstream historical accounts - has in actual fact been part and parcel of any and all human culture and civilization. We encounter here the author's commitment to material practices as foundational to any and all forms of abstraction.

In the book's first part, "The Industrial Age", the author first retraces Babbage's writings to argue that his theories show that industrial machinery arose out of historically existing divisions of labor. That is, machines do not determine how labor ought to be divided; but the historically forming divisions of labor more crucially shape the emergence of machines. The next chapter on "The Machinery Question" mobilizes Marxian theories that similarly argued that technology emerges as a result of the division of labor, not the other way around. Interestingly, here we also encounter the idea (by Marx) that an extended division of labor, not the scientific enterprise, must be seen as the inventor of machinery. In other words, tool-makers and machine operators were constantly contributing to the invention of new machines (p. 90). In the book's final chapter, we also meet the intriguing argument that the steam engine was as influential as it was not due to its intrinsic superiority, but due to its corresponding most aptly to the needs of capital at the time. In sum, one might say that the book's first part presents a history of the industrial age that explains its development not by recourse to technologies as such, but by looking at the given social divisions of labor at the time and how these relate to the emergence of technology.

The book's second part, on "The Information Age", starts with an interesting chapter on cybernetics and its emergent focus on self-organizing systems. We see here the early seeds of today's self-adapting machine learning systems. McCulloch & Pitts in 1943 famously introduced the notion of artificial neural networks which, as Pasquinelli argues, were not biomorphic but instead as a technique for the self-organization of information (p. 135). Crucially, this self-organization of artificial neural networks must be understood as occurring in relation to the external world. Pasquinelli retraces how cyberneticians' claim to the imitation of living beings revealed more about the society and labour relations of their time than nature per sé. Here a nice crystallization of the basic notion undergirding historical epistemology: "the organisation of labour in a given epoch influences the formation of technologies and instruments, and thereafter of scientific paradigms, conceptions of nature, and models of the mind too." (p. 154). This leads to Pasquinelli's important insight that information technologies became as powerful as they are today because of their "inborn capacity to capture social cooperation" (p. 154) - also the reason why connectionist AI with its statistical and inferential algorithms won out over symbolic AI. The project of AI, put simply, is sociomorphic and not biomorphic. In chapter 7, we learn about how von Neumann's approach differed from that of other cyberneticians - and what were his strengths. Chapter 8, surprisingly, turns to an anylsis of Friedrich Hayek's "The Sensory Order" (1952) - a book that is seen as a milestone in the genealogy of connectionism. Hayek, among other things, elevanted the role of classification and pattern recognition as crucial tasks of the human mind. The book's final empirical chapter traces the 1958 invention of the Perceptron - a statistical neural network for pattern recognition- by Frank Rosenblatt. This is also where psychometrics and statistics took their hold in the field of AI as they crucially influenced Rosenblatt's work.

The book's final chapter crystallizes the author's main argument once more: AI is a project of imitating the social division of labour (or a "social intelligence), not of imitating a biological intelligence that is somehow hidden in human brains. Interestingly, machine learning represents for the author the "automation of automation" (p. 248). This time, he further concludes, we are seeing the automation of management rather than the automation of labour - management is algorithmically replaced, precarious jobs are multiplied. How can this process be subverted? Not by encoding hard ethical rules into AI systems, because this fails to tackle the deeper political economy of AI. Instead, Pasquinelli urges that "The first step of technopolitics is not technological but political. It is about emancipating and decolonising, if not abolishing as a whole, the organisation of labor and social relations on which complex technical systems, industrial robots, and social algorithms are based - specifically, their inbuilt wage system, property rights, and identity politics" (p. 253). Critique of technology, put simply, has to start with politics, not with technology.

I greatly enjoyed this book not only because of its counter-intuitive theoretical and methodological approach - and the resulting argument that AI is sociomorphic rather than biomorphic - but also because of its meticulous reading of some of the core theorists of both the industrial (Babbage, Smith, Marx) and the information age (McCulloch & Pitts, von Neumann, Rosenblatt). Having these texts connected to their social contexts and the prevailing social divisions of labor was very interesting and enlightening. The author's main claims, thus, are definitely plausible. I would have wished for a deeper elaboration of his theoretical-methodological approach, though. Why did he focus on these particular texts, and not others? Why did he choose a historical epistemology, rather than, say, a Foucauldian genealogy? What exactly did he mean towards the end of the book when he suggests the first step of critical analysis ot be "deconnectionism" (p. 251)? All these are not weaknesses per sé, I'm just saying that I would have greatly enjoyed a deeper discussion of these. Nonetheless, this book clearly deserves 5 stars.

[Ben · Rating 4 out of 5]

An impressive re-contextualization of what's currently hyped as "artificial intelligence." Perhaps surprisingly, the book is as much an argument over the philosophical status of mathematics and scientific innovation, generally speaking, as it is an intervention into what is commonly labeled "AI" nowadays. Matteo P. walks a fine line, but makes a strong case for a constructivist interpretation of math and science while also avoiding falling into technological determinism. (Humans' social use of tools just happens to both suggest and constrain subsequent stages of inquiry — which the tools' own uses can't but enable.) Although some chapters are more technical than others, the main argument is that current machine-learning, neural-network versions of "AI" should primarily be seen as extensions of the long history of (social) labor automation and expropriation (and, of course, exploitation) — not inherently, or primarily, to do with consciousness per se at all. His summary of the "symbolic" (mind-centered, "top-down") vs. experimental (neural-net, machine learning, or "bottom-up") approaches to "AI" advances ring true in this regard. Any "intelligence" in "artificial intelligence" is expropriated from the social activity of human beings, on the data end as well as on the modeling end.
One minor criticism of the book is that, if reading chapters sequentially, one sometimes wishes that the chapters were more explicitly linked together. Even though the introduction and conclusion do tie things together, insights established within each chapter — and/or some summaries — could have been more frequently referenced within other chapters.
Some interesting asides also pop up throughout the exposition. For example, Matteo P. highlights the contingency of technological and engineering advancements by mentioning a 1969 "hack" of an early, low-res. machine learning pattern recognition approach. The "hack" of an ambiguous pixelated image within the 20 x 20 grid was designed to show its limitations, which "derail[ed] military funding and neural networks research until the late 1980s" (even though the limitations highlighted at the time were a function of the limited computing power available, not the underlying method per se). And this 20-year derailment was to the branch of AI-related research that eventually bore fruit in recent large language models like ChatGPT. There's also some brief discussion of functionalism versus von Neumann's sense of operative models independent of their "material" substrate implementation, and first-order versus higher-order representations of equivalent processes.
For Marxist philosophy nerds, there's some interesting commentary on Marx's "Fragment on Machines," and its debated implications for either the eventual downfall of capitalism altogether or just more mundane implications for relative surplus values.
At a minimum, definitely worth reading, if only to be able to debunk all your ChatGPT-enthusiast friends who start thinking they're chatting with a sentient entity. And, at its best, provides a valuable philosophical method for assessing further developments in what will no doubt continue to get hyped as "artificial intelligence."

[naisokram · Rating 4 out of 5]

I picked up this book intrigued by its promise to explore AI’s development not through the lens of technology and science, but through the history of labor, economics, and society. The first chapter reignited my love for math, the second drew me into deep philosophical reflections on knowledge, and the rest of the book challenged me - sometimes stretching beyond my comfort zone as a non-expert. While parts delve into professional-level discussions, the core ideas are too compelling to ignore.

Here’s what I grasped from this book:

1.AI is not magic
Pasquinelli dismantles the myth of AI as a mystical force, arguing instead that it emerges from collective human intelligence, shaped by economic and social structures. AI isn’t just brilliant innovation, it’s a product of labor dynamics and societal needs.

2.The labor theory of automation
The book ties AI to the division and management of work, much like machines in the Industrial Revolution. Rather than replacing laborers outright, AI often automates managerial oversight - the “eye of the master” - while still relying on hidden human labor (data annotators, content moderators, etc.).

3.Surveillance and control
The title references Marx’s idea of the “master’s eye” - the surveilling gaze over workers. Today, algorithms and data systems enforce workplace discipline, embedding historical biases (like gender discrimination) into AI. This exposes how AI isn’t neutral; it mirrors and amplifies societal power imbalances.

4.Resistance and reimagining
Pasquinelli emphasizes that technology is political. The final chapter rejects the idea of an inevitable AI future, urging readers to ask: Who controls these systems, and who benefits?He argues that real change requires collective action - demanding transparency, equitable automation, and public-oriented technology.

This book is a call to action: understand AI as a social process, resist uncritical hype, and fight for a future where technology serves the many, not the few.

[Rahel · Rating 3 out of 5]

A very interesting book that at times got a bit too caught up in the specificities of what it was trying to tell the reader. Pasquinelli excels at chronicling the historical developments of AI - the first and last third in particular are quite engaging, while the middle, as well as two specific chapters, get a bit lost in the theoretical sauce and prove to be far less accessible. I work in the field of computational social sciences and am thus no stranger to convoluted and dense texts, but I found myself struggling to keep up with the chapters focused mainly on labor and economic theory, as well as philosophy, fields that I am most definitely uninitiated on. The two aforementioned chapters, which fall into this category, were both previously published in other academic publications, and I feel that that was quite evident, in terms of focus, depth and writing style.
Nonetheless, I found the author's arguments to be convincing and am left with loads of thoughts I want to continue thinking through. His perspective on AI as the "[imitation of] the intelligence of labour and social relations", rather than "biological intelligence" links up wonderfully with other proposals of AI behavior as a representation of human values and the interactions (see The Good Robot on that one) and thus provides necessary further dimensions to ongoing discussions on automation and (artificial) intelligence

[Pablo Benavides · Rating 5 out of 5]

I had the opportunity to read this book a few months ago. Now that I’ve just joined Goodreads, I felt it was the perfect time to review my notes and reflect on the topic. This book guides us through the premise that work and labour are specifically social processes, involving dynamic interests and hierarchies within the society we live in.

It begins with theoretical mathematical developments related to cultures that connected these theories with cults and religiosity. After that, we revisit the development of computation—how Babbage developed machines that laid the foundation for our current approaches in computer science, and how, from the beginning, these were intended to impact existing work relations. The author then introduces Marx and explores the “machinery question,” its historical impact and reflection, and moves toward how labour evolved alongside technological changes.

What surprised me most was the modern take on how technology ultimately leads to AI. The people involved are impressive. Even acknowledging Hayek’s contribution to the overall discussion on human thought—and how his position later converged into neoliberal theory—is striking.

If you want an in-depth introduction to the history of labour and technology, along with a fresh reflection on where we are heading, this book has it all.

[John · Rating 5 out of 5]

Η αναγνώριση προτύπων αποτελεί την αυτοματοποίηση μιας «υψηλής κερδοσκοπικής ικανότητας» – την ερμηνεία μιας εικόνας ή ενός περιβάλλοντος.
Συνδέεται με την αυτοματοποίηση της «τοποθετημένης γνώσης» που αποτελεί μέρος της πολιτιστικής κληρονομιάς, συμπεριλαμβάνοντας σιωπηρή και ρητή τεχνογνωσία, ταξινομίες και γραμματικές.
Η αναγνώριση προτύπων, σε συνδυασμό με την ψυχομετρία, ενισχύει κοινωνικές ιεραρχίες, ταξικές, έμφυλες και φυλετικές προκαταλήψεις. Εμμέσως δηλώνει τι μπορεί να αυτοματοποιηθεί και τι όχι, επιβάλλοντας νέες μετρικές νοημοσύνης.
Η άνοδος της αναγνώρισης προτύπων συνδέεται με την «επανάσταση του ελέγχου» και την ανάγκη διαχείρισης πολύπλοκων οικονομικών και κοινωνικών συστημάτων, επεκτείνοντας την επίβλεψη από την εργασία σε συλλογικές συμπεριφορές.
Ο «κυρίαρχος αλγόριθμος» της AI, ειδικά η «τεχνητή γενική νοημοσύνη», περιγράφεται ως μια «στατιστική ψευδαίσθηση που προβάλλεται από δεδομένα», αποτελώντας τελικά μια «πρωταρχική εκτεταμένη κοινωνική μορφή».
Εν κατακλείδι, η Αναγνώριση προτύπων αποτελεί ένα θεμελιώδες πεδίο της AI, με βαθιές ρίζες στην ιστορία της επιστήμης και της τεχνολογίας, και σημαντικές επιπτώσεις στις κοινωνικές δομές και σχέσεις.
Α ΜUST READ

[Chelsea · Rating 5 out of 5]

A stellar critique of the utopian and regressive promises of technology, ranging in its historical-theoretical analysis from Marx to Lefebvre to Deleuze, Pasquinelli's book traces a genealogy of the division of social labor and its metrics to argue that this is the operative principle of ‘Artificial Intelligence’, showing how “Marx’s theory of automation and relative surplus-value extraction share common postulates with the first projects of machine intelligence” (239). This masterwork is hardly 250 pages yet brimming with insights on the social materiality of the abstractions contained in algorithmic governance, and above all the 'general intellect,' or the possibilities of complex self-organization contained in such an algorithmic bureaucracy.
What philosophers of the 20th century had missed about cybernetics, Pasquinelli shows to be crucial to a speculative and scientific view of the stores of social labor humming in the global nervous system of capital. The breadth and depth of this book is breathtaking, and all those who long for an end to unnecessary suffering should read it.

[Drew · Rating 5 out of 5]

4.5 stars

The Eye of the Master by Matteo Pasquinelli is a book that challenges the myths about artificial intelligence (AI). The author argues that AI is not a quest to imitate biological intelligence, but a continuation of the automation of labor and social relations. He shows how AI is shaped by the industrial factories and the collective behavior of workers and consumers. He also exposes the political implications of AI, such as the extraction of data and value, the reproduction of social inequalities, and the threat of mass unemployment. The book is well-researched and engaging, offering a critical and alternative view of AI. The author calls for a new literacy and awareness of AI that can empower people to challenge the dominant paradigms and to imagine new futures.

The only real minus for me with regards to this book (not counting its denseness) is that the history they shed light on doesn't really go past the middle of the 20th century. I would have liked to have seen more about the evolution of AI in our contemporary times.

[Murat · Rating 5 out of 5]

This an extremely dense book of history of science (difficult to read quickly) that however retraces, very precisely, the roots of the current AI paradigm, Deep Learning, especially in the last chapters. It reaches the years in which the first neural network (the Perceptron) was invented, 1950s, and stops there basically. Highly recommended if someone wants to know where really Deep Learning (not just Good Old Fashioned AI) is coming from. It's very accurate about historical details and tries to explain the mathematics of deep learning in simple terms (multivariable analysis, which remains difficult to an average reader). It's a technical history that is claimed as a social and economic history, not sure all engineers will be happy be studied under the coordinates of political economy. In short, the most complete history of a complex matter such as AI algorithms I ever read.

[KT]

oof lol i need to reread this when i own a copy + can write all over it. (or get an ebook i guess. im limited by my library book!!) the beginning was especially interesting and strong; the discussions abt babbage + marx were fascinating. the hayek chapter didn't make a lot of sense to me in terms of why it was included, BUT that's something i'd just aim to go back and read again.

overall quite interesting, not to mention refreshing (so much of tech capitalism critique isn't marxist in nature).

side note, i found out about this book because i read this article by edward ongweso jr and was curious. great article btw! i recommend!

[Andrew · Rating 1 out of 5]

Really didn't love this book. I'd say this was 3 separate books. Book 1: the introduction and conclusion, which should be read independently from the the rest of the book, seeing as they had very little to do with the content. Book 2: was the beginning chapters about division of labor. It seemed like the author was taking the opportunity to make some points he finds very relevant on the genesis of a lot of Marx's content. It was interesting, but had nothing to do with AI. Book 3: A short, high level history of the origination of neural networks which was interesting, but there are no doubt better histories on the subject. Overall. not my favorite book.

[Ruairí]

More of a genealogy/intellectual history than a social one, but definitely puts social relations in the analytical driving seat. Most of the maths/cybernetics was over my head but I got a lot out of it.

His point is that the history of AI as we know it today is one of ever more precise measurement and division of labour. AI isn’t a robot replacing individual workers, but the automation of “the eye of the master” i.e. management. However in doing so, AI also requires forms of labour that can be managed by these systems - the archetypal precarious platform worker.

Tracing AI back to the jacquard loom rather than very early ‘computers’ illuminates this point.

[Victor Ogungbamigbe · Rating 5 out of 5]

Incredibly well researched and developed criticism of the techno-fabulism of modern AI. Pasquinelli creates a critical historiography of AI based on the labor theory of the machine, connecting the works of Babbage and Marx before drawing on the distinctions made between symbolic and connectionist ideologues of the early days of machine learning. Very enjoyable, though highly complex read, but that can be forgiven given the subject matter and sheer breadth and depth of topics covered. Also, the cover looks fucking fantastic and that has to count for something.

[Kathleen O'Neal · Rating 3 out of 5]

This book was assigned by Dr. Denise Albanese as part of the Cultural Study of Science and Technology course I took in the spring of 2024. In this work, Matteo Pasquinelli places labor at the center of the historical development of artificial intelligence technologies. Particularly interesting to me was Pasquinelli's analysis of the ideas of Friedrich Hayek in this context. In this work, Pasquinelli centers the concept of the algorithm and seeks to define it in broader social terms than the strictly technological terms that most of us are familiar with in reference to the word.

[LustHeroin · Rating 2 out of 5]

Health to the translator.
Half of the book is propaganda of Marxist leftist theory and an attempt to consider mathematical models as slave labor that will replace man. That is, the works of postmodernists in the era of the information revolution. The book should be placed on a par with "leak", "dreadnought", and "bunga-bunga". For the first time, popular science turned out to be more difficult for me than scientific works. It is written very dryly, and this is not the fault of the translator. The problem with the whole work is the application of post-Marxist rhetoric to perceptrons or Babidge's theory, and considering the same in the information age is already difficult for the author. After all, doing a retrospective is more difficult than an analysis.and author failed it.

[Mtume Gant · Rating 5 out of 5]

An absolutely essential text. What the history of technology has lacked is a historical materialist approach and the author takes it in his approach to A.I. We have to accept that our tech development is directly intertwined with the capitalist project of Western domination through labor exploitation and extraction, why A.I would be any different is lying to oneself and living in the idealist myth that tech development is a history of great individuals and not a social project looking to automate the collective nature of labor.

It’s a brilliant account here. No notes. Read it.

[Ehab mohamed · Rating 3 out of 5]

قرأته في نسخته العربية الصادرة عن دار صفحة سبعة باسم : عين السيد.

يمكن القول بأن الكتاب تفسير ماركسي لنشأة الذكاء الإصطناعي باعتباره ناشئ نتيجة لسياق اجتماعي أوسع من أدوات وعلا��ات الإنتاج وأنه لا يزيد عن أتمتة تقسيم العمل وإدارة العمال.

الكتاب كما هو واضح من عنوانه تاريخ اجتماعي أما تاريخ التطور التقني فليس هدفه وهو ما جعله يورد دقائق تقنية لا هو شرحها بتفصيل فنفهمها، ولا هو أوردها بإيجاز فنأخذها على عمومها، ولكنه أورد أنصاف شروح، وأنصاف تفسيرات لأمور غاية في التقنية والتعقيد، فلا فهمناها ولا تخطيناها.

[Jon · Rating 4 out of 5]

This academic work traces the evolution of the technologies of the automation of knowledge work from Babbage to Marx to Hayek (unexpectedly for me) to the AI of the present. The general point is that technodeterminism overstates technology's impact on the means of production and rather society informed technological advancement, encoding its norms into the technology itself. Of course the reality is a bit of both, but it's a worthwhile corrective to a moment where AI is positioned as somehow magical.

[Owlseyes]

https://www.versobooks.com/en-gb/blog...

[Brian Swain · Rating 3 out of 5]

Interesting read if you're very close to the AI field. Extremely dense and obtuse writing style, i.e., it reads like a doctoral dissertation. If this isn't your style, go for something a bit more layman-esque.