An investigative journey into the sources of morality in artificial intelligence and how this impacts our society.
Offering a rigorous, fresh perspective on how technology has brought us to this place, Mechanisms of Morality challenges the long-held assumption that technology is an apolitical and amoral force. Having worked in the field of artificial intelligence for over 30 years, Smith reveals the mounting evidence that the mechanical actors in our lives have, or at least express, morals: they're just not the morals of the progressive modern society that we imagined we were moving towards. Instead, as we are just beginning to see--in the US elections and Brexit to name but a few--there are increasing incidences of machine bigotry, greed and the crass manipulation of our basest instincts.
It is easy to assume that these are the result of programmer prejudices or the product of dark forces manipulating the masses through the network of the Internet. But what if there is something more fundamental and explicitly mechanical at play, something inherent within technology itself?
This book demonstrates how non-scientific ideas have been encoded into our technological infrastructure. Shedding light on little-known historical stories and investigating the complex connections between scientific philosophy, institutional prejudice and technology, this book offers a new and more truly scientific vision of ourselves.
This is a great book by University College London professor and vastly experienced AI and algorithm expert Rob Smith. It's marketed as being about the way algorithms polarise online communities and encourage division, but it's much more than that - history, science, mathematics, statistics and personal anecdote covering vast swathes of territory.
Smith's main skill as a scientist, thinker and writer is in mining out the nuances and ambiguities of scientific development - no scientific idealist. He has a high regard for what makes us human and the way we transcend machines and code.
I won't try and summarise the main gist of the book, but I do think, given the prevalence of algorithms and forms of 'AI' (which Smith says should be better called 'pseudo intelligence') in our online lives (for example, they form the backend of social media and search, curating the material you see), we should be informed about their function and influence, and never assume that we're impervious to that influence. There's a high probability we have all been (and are) influenced by algorithms, maybe even to the extent of shaping our perspectives.
Here's my takeaway / two cents worth:
Given the tendancy of algorithms to polarise in the online communities that we use, and their sub-human operating methods, what is required of us is to be conscious of our vulnerability and employ our uniquely human faculties (those things which are not a feature of algorithms) - such as an appreciation of creativity, nuance and diversity, and the exercising of empathy and compassion - so as to retain our humanity, equilibrium and health. In this way we choose not to acquiesce to polarisation, and subvert the simplistic and reductionist modes or ecosystems that favour algorithmic systems.
If, as Marshall McLuhan and Neil Postman suggested in earlier media epochs, media shapes how we think (not just what we think), then this kind of resistance to algorithmic systems is also protecting the wiring of our minds. The world would be a much less colourful place if our consciousness functioned in the way an algorithm does. The concern, I think, is not so much that AI will catch up, but that we will meet it in the middle by being reduced.
I want to give this a higher rating because the topic is so interesting, but the mathematical understanding necessary to read this book will make it inaccessible to so many people. It's been a long time since my last calculus class, so a lit of this went over my head. But if you skip over the equations, there is so much to think about! I hope one of my friends reads this so we can discuss!
This book takes a little reading. Let me explain that sentence. This book is written by a computer engineer so he is very attracted to numbers and formulas. There are a lot of these in this book. However when you get past all of it the message is a very good one on human interaction and bias.
I’m on the fence with this one. The subject matter is of interest to me, but the proto-dystopian view of data analytics is... difficult to parse. I don’t believe that what the author posits is untrue, but it does feel weighty on the side of unintended evils in technology.
i didnt expect to like this much at all, especially with its title / cover but the book picks up at the feminist section at chapter 7 and is pretty solid for the rest
Siapapun yang mengamati Facebook saat waktunya pemilu, entah di Indonesia ataupun Amerika, bisa melihat jelasnya jurang pemisah antara pendukung si A dan si B. Media sosial menjadi tempat bertengkar sengit, yang makin lama jurangnya semakin lebar, merembet dari dunia maya ke dunia nyata, kadang punya efek yang fatal.
Menurut Robert Elliott Smith, seorang pakar algoritma evolusioner, dengan pengalaman 30 tahun berkecimpung di dunia AI (artificial intelligence), hal seperti itu menyedihkan tetapi sudah dapat diduga, akibat algoritma yang diterapkan di media sosial. Bukan hanya Facebook, bukan hanya urusan politik, tetapi hal ini sudah mendarah daging di berbagai urusan manusia yang semakin terkomputerisasi.
Bagaimana hal ini bisa terjadi, sedangkan para pionir teknologi awalnya memimpikan Internet sebagai alat untuk membangun masyarakat yang lebih adil, terbuka dan toleran?
Menurut Smith, pada intinya adalah karena algoritma selalu berupa penyederhanaan, mereduksi dunia nyata yang kompleks menjadi kategori dan nilai-nilai yang terukur untuk membuat generalisasi yang dapat ia gunakan untuk menjalankan programnya. Begitu pulalah bagaimana prejudice bekerja: dengan mereduksi hal kompleks menjadi suatu pandangan sempit.
Dalam buku ini Rob Smith membangun argumennya dengan menerangkan secara sangat (sangat!) detil dan mendalam tentang bagaimana algoritma bekerja, teori-teori apa yang mendasarinya, bagaimana sejarah kemunculan teori-teori tersebut, dalam konteks apa mereka lahir dan bagaimana mereka dipakai, sehingga mengakibatkan munculnya efek negatif seperti ketidakadilan perlakuan terhadap minoritas dalam sistem komputerisasi (seperti di perbankan, kepolisian, image recognition, dll), dan permusuhan yang tajam antara kubu A dan B di media sosial untuk isu-isu tertentu.
Teknologi tidak muncul dari vakum, dia dibuat dan dikembangkan dalam konteks sosial budaya dan filosofi kemasyarakatan tertentu. Karenanya, dalam teknologi selalu ada bias. Mengenai bias dalam dunia keilmuan, dan hubungannya dengan algoritma ini, Smith memberikan banyak contoh.
Misalnya di akhir abad 19, di masa kemunculan teori Darwin tentang evolusi dan keragaman random di alam, Herbert Spencer, seorang sosiolog, menciptakan jargon 'survival of the fittest' untuk teori tersebut dan menafsirkannya dalam konteks sosial, bahwa ada suatu kondisi ideal (the fittest), dan yang selain itu adalah 'kurang ideal' atau bahkan 'tidak normal'.
Ide ini mendorong ilmuwan lain, Francis Galton, menelurkan teori eugenics, bahwa masyarakat ideal bisa dicapai dengan menaikkan jumlah orang dengan "kualitas" tinggi, dan menyetop pertambahan orang ber"kualitas" rendah. Lalu muncul test IQ, yang awalnya diciptakan untuk mengenali anak-anak yang kesulitan belajar dan membutuhkan bantuan, tetapi ke depannya malah dipakai untuk kategorisasi siapa yang cerdas dan siapa yang tidak.
Eugenics dan pengategorian kecerdasan melalui test IQ, yang keduanya dianggap ilmu valid pada masa itu, berkaitan erat dengan ide-ide supremasi ras, dan memunculkan hal-hal negatif yang mengerikan, seperti pemaksaan pemandulan perempuan yang dianggap berkualitas rendah (miskin, bodoh, IQ rendah, yang dianggap akibat genetik, bukan sosio-ekonomi), bahkan eugenics ini pula yang mendorong ideologi Nazi yang berujung holocaust dan perang dunia II.
Ya, ilmu yang pada saat itu dianggap valid, ternyata bisa salah fatal. Dan hal itu bisa terjadi jika mereka yang berkecimpung di dalamnya (yaitu para ilmuwan) tidak mengakui adanya bias, dan tidak mengevaluasi diri dan ilmunya secara terus menerus.
Masih soal bias ilmu, buku ini juga membahas secara dalam tentang konteks lahirnya teori-teori ekonomi, yang kadang demi optimasi dan profit, tidak mempedulikan manusia-manusia yang berkaitan dengan prosesnya, memunculkan algoritma yang tidak manusiawi. Contohnya kasus yang mencuat tentang protes para pekerja gudang Amazon, yang mengekspos suasana kerja yang tidak manusiawi, segala gerak-gerik mereka dimonitor, bahkan istirahat dan ke toilet hanya diberi waktu sangat sedikit, semua demi optimasi.
Lalu juga dibahas tentang ilmu statistik, "bell curve" kurva distribusi error yang sangat berkaitan dengan ide 'normal-tidak normal' (seperti tafsiran Herbert Spencer di atas), seringkali mengakibatkan nilai-nilai 'outlier' dihilangkan, padahal menurut Smith, justru nilai-nilai outlier ini bisa menjadi pelajaran penting dalam evaluasi suatu algoritma. Dibahas juga tentang konteks sosial ekonomi yang mengakibatkan kaum perempuan dan ras minoritas jumlahnya tidak banyak di dunia ilmu komputer. Kesemua bias ini sangat berkaitan dengan cara kerja algoritma. Bias yang bertumpuk-tumpuk di setiap langkah algoritma, bisa menghasilkan efek negatif yang tidak diinginkan.
Mengenai artificial intelligence, Smith mengingatkan, pada dasarnya algoritma adalah model, representasi sederhana dari dunia nyata yang kompleks. AI dianggap menyerupai intelegensi manusia, atau sebaliknya, otak manusia dianggap seperti komputer saja. Benarkah? Smith mengajak pembaca membandingkan kerja komputer/algoritma dengan cara manusia berpikir, yang menurutnya sangat berbeda. Ia juga membahas ilmu linguistik dari Noam Chomsky, dan menyoroti kontrasnya arti kata-kata 'informasi' dan 'komunikasi' di dunia ilmu komputer dan di dunia nyata. Algoritma tidak mengerti tentang 'makna' atau semantik, ia hanya peduli dengan sintaks dan grammar, serta probabilitas untuk optimasi keluaran. Sementara manusia memahami makna suatu kata, gambar, atau informasi secara luas, tersurat dan tersirat, dari sangat jelas hingga kesan terhalus, melalui intuisi dan pengalaman pribadi yang beragam, yang tidak bisa ditangkap oleh algoritma. Keragaman ini, adalah kekayaan manusia, yang membedakannya dengan mesin.
Reduksi keragaman inilah yang terjadi di komputer, internet dan media sosial, ditambah dengan faktor ekonomi dan optimasi target, mengakibatkan terjadinya pemisahan konsumen media sosial ke dalam kategori-kategori sempit untuk kepenting profit dan politik. Mendorong pihak-pihak dengan perspektif A mengumpul di satu sisi dan perspektif B di sisi lain, memunculkan echo chamber yang semakin memperkuat kubu-kubuan. Ketika pemisahan ini mencapai equilibrium, masing-masing pihak sudah punya perspektif permanen yang sulit bahkan mungkin tidak bisa dihilangkan. Menyedihkan bukan?
Adakah solusinya?
Menurut Smith, solusinya tidak mudah, tetapi kita harus mencobanya, kalau tidak mau efeknya semakin parah. Berkali-kali Smith mengingatkan bahwa model algoritma berangkat dari 'jendela' sudut pandang tertentu dalam memandang problem yang ingin diwakilinya, dan jendela tersebut sangat dipengaruhi bias sudut pandang programmernya.
Smith mengajukan beberapa solusi yang bisa diambil: Kita harus mempertimbangkan kembali model algoritma yang dipakai dan meneliti penyederhanaannya, coba pindahkan 'jendela' tadi untuk menemukan perspektif baru. Kita harus bertanggung jawab atas efek yang muncul akibat aplikasi algoritma terhadap kehidupan manusia. Kita harus akui adanya bias, dan mengatasinya. Bagaimana? Salah satunya dengan diversifikasi praktisi perancangan pemrograman yang bersangkutan. Bukan artinya programmer yang sekarang buruk atau rasis, tapi bisa jadi mereka tidak mengerti perspektif dari ras/gender/budaya lain yang berkaitan dengan dunia nyata yang diwakili oleh alogritma mereka.
Dia kemudian menceritakan studi algoritma evolusioner tentang sistem imun manusia, yang ternyata memunculkan keragaman keluaran, tidak hanya satu 'the fittest'. Ternyata di alam, keragaman itulah salah satu bahan penting dalam evolusi. Dan menurut Smith, itu juga yang harus diterapkan dalam AI: menjaga keseimbangan antara keragaman dan optimasi, supaya bisa berinovasi secara efektif. Hal ini bisa didapat dengan cara menerapkan 'fitness sharing' dalam rancangan algoritma, menjaga supaya tidak ada satu agen/individu yang mendominasi terlalu kuat sehingga menghancurkan keragaman.
Smith menyatakan kekhawatirannya jika dunia AI terlalu silau dengan kemajuan teknologi tetapi kehilangan rasa kemanusiaan. Bagi generasi muda praktisi AI, mungkin pandangan ini dianggap 'nggak asyik'. "Party pooper" katanya, menceritakan pengalamannya di suatu konferensi Google yang didominasi anak-anak muda. Tapi saya melihatnya sebagai kebijaksanaan seorang praktisi senior, yang telah melihat efek negatif algoritma tetapi juga punya informasi tentang cara mengatasinya, yang kalau kita mau, bisa dilakukan. Ia seperti ingin mengatakan, "Silakan berinovasi, tapi hati-hati, jangan sampai menyakiti. Belajarlah dari masa lalu."
Bahasan buku ini sangat luas dan mendalam, sulit bagi saya mereduksinya menjadi model yang simpel dalam satu review pendek.
Bagusnya para ilmuwan bidang informatika dan praktisi dunia artificial intelligence baca bukunya langsung deh.
Dr. Robert Elliott Smith has many cautionary tales to tell about the simplifications of human nature throughout history by scientists, economists, educators, policymakers and, most recently, computer programmers. That’s particularly interesting because Dr. Smith (University College London) is himself a computer scientist and expert in AI (artificial intelligence). And it is his own field of expertise that potentially threatens so many professions in the near future while also polarizing civil society into angry, dismissive factions in the present moment.
Dr. Smith feels that the ability for AI to effectively replace human intelligence and creativity may, like the mechanizations of the past, be overstated. He knows the value of AI in optimizing the shape of an airplane wing, having engineered such algorithms himself. But he also knows that human interaction and adaptation is more complex than most AI programmers recognize. And he bolsters that observation with both serious analysis and anecdotes from interactions with colleagues in his field, as well as an interesting homage to Salvador Dali, among other creative minds.
But Dr. Smith is greatly concerned about the inherent prejudice of algorithms on civil society and democratic institutions. Queried on this concern in our Purple Principle interview, he argued strenuously and eloquently for regulations to be placed on major social media platforms, such as Facebook, to prevent the Cambridge Analytica-style scandals of the future.
Rage Inside the Machine is, in part, an expansive history of science journey. Newton, Galileo, Malthus, Darwin, Babbage, and Adam Smith, among others, are part of this narrative. But it is also one part personal memoir of the civil rights era. Having been raised and educated in Birmingham, Alabama, at the time of enforced desegregation, Smith knows something about human tribalism. He emerged from those experiences with a great respect for diversity, not simply from a moral standpoint; but as an essential ingredient for a modern, prosperous democracy.
For those with less technical backgrounds, these personal reflections may be the most interesting sections of the book. For those with strong science and technical interests, Smith’s facility with a wide range of subjects will be impressive and engrossing. But all readers will come away with a renewed respect for the uniqueness of human intelligence, both that of Dr. Smith as a wide-ranging author, and of our own, as a creative and adaptive species.
Worth Further Thought (excerpt from page 10): “But reductionism and simplification when applied to people and society as a whole have a dangerous, natural tendency to lend support to bigotry. It has taken centuries of human effort to continuously overcome these tendencies and work towards a more progressive and tolerant society. However, simplifying modelling has now become an autonomous part of algorithms in our global infrastructure…”
Rage Inside the Machine is a delectable blend of (what I like to call) armchair maths, history, autobiography, current events (and what led to them), cautionary tale, and call to action. As a computer professional, for lack of better parlance, this really spoke to me. The book is well-written, easy to follow, and balances hard science with soft entertainment, expertly. I highly recommend this to anyone whose life is affected by algorithms, that is to say, everyone. I plan to get a hard copy to reread at least a few times, as well as pursue some of the bibliography. Thanks to NetGalley and the publisher for the eARC in exchange for an honest review.
Makes a thought provoking argument about how we should frame our relationship to algorithms and so-called "artificial intelligence".
Money quote for me: "fundamentally, it is because algorithms always embody simplifications, and simplification is always at the core of prejudice."
The book expands on this using a surprisingly broad swath of history, clear explanations of current algorithms and research, and some convincing philosophical arguments.
Foremost a history lesson on algorithms and computing, Rage Inside the Machine examines the deep roots of the prejudices unwittingly baked into the programmes that we interact with every day. See my full review at https://inquisitivebiologist.com/2019...
This book should be required reading for anyone working in tech alongside Algorithms of Oppression: How Search Engines Reinforce Racism by Safiya Noble, Technically Wrong: Sexist Apps, Biased Algorithms, and Other Threats of Toxic Tech by Sara Wachter-Boettcher, and Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy by Cathy O'Neil.
Not a lot of information on how to prevent AI from being discriminatory from either an end user or a programming perspective. Interesting read for anyone not in the field but ultimately does not deliver on the tagline.
Felt like I was reading a fun university lecture on real world issues. Although the computer science jargon was a bit much, and some of the explanations were over detailed…it was still interesting to learn about. Yippee. Algorithms are so scary.
A must read 'AI is the hole, not the doughnut' A useful model or technology isn't a replacement for the complexity and diversity of humanity but a tool for a better understanding
Not as advertised. Half math text and one third professional memoir with just the slightest glaze of analysis and virtually no recommendations for how to stop the negative impact of algorithms.
A thorough development of ideas that lead to our current model of AI, reframing AI as Pseudo intelligence and identifying root causes for bias and polarisation in systems.
This was a really well done blend of memoir and explainer about algorithms and how they're tuned especially to elicit strong reactions, and how we can maybe fix that in the future. At the time I'm reading this a good chunk of these predictions have come true, so it's kind of darkly funny to read this now.
This book has a lot of information densely packed. It's a comprehensive history of algorithms, pre-computer age through the present day.
My TLDR is that algorithms are not so smart. We're ceding too much control to calculators that deal with the primary objectives of probability and profit generation.
Book is interesting and a very useful warning against trusting algorithms too much. Unfortunately the writer can't resist the temptation to talk too much. Even his end warning is spoiled by giving too much details. A very strict editor would have made a better book. 2 stars lost for giving too much information.