Artificial Unintelligence: How Computers Misunderstand the World (Mit Press)

Clear, insightful must-read explainer on Artificial Intelligence and the social implications. Anyone interested in what Google AI lab head Fei-Fei Li has called "human-centered AI" should read this book. Broussard, a professor of AI and journalism at NYU, makes the argument that the narrow AI that is currently being developed and deployed is far from the dreams of science fiction of general AI and that gap influences what society thinks our data-driven systems are capable of. The result is that the enormous opportunity and possibilities for the future of AI are masking today's real-world problems of implementation: bias, representation, and real-world problem solving. Broussard has a journalist's keen eye for detail and gift for story-telling. The result is a readable and careful summary of the current academic research on the social, legal and cultural impact of AI for the general public with an urgent call to action to design and implement technologies that benefit all of society.

This book may easily win the prize for the best book that no one who should read it wants to. And that is precisely why it is the one book that everyone must read.Meredith Broussard is a coder, educator, and a computational journalist that specializes in algorithmic accountability reporting. Which is to say that she is a very tech-savvy investigative journalist that emphasizes statistical analysis. (The algorithmic and computational side of it is method more than purpose, to my way of thinking.)She is most definitely not a Luddite. “My goal in this book is to empower people around technology.” She embraces technology and the power of algorithms, but with a caveat. She is wary of the autonomous school of computing who wants to turn it all over to the machines. And she makes a very strong case that doing so is both impractical, in the foreseeable future, and inappropriate. “We need to stop fetishizing tech.”It’s an important message. In my own words, machines will never think in the human sense because thought is relative. Even humans have difficulty interpreting reality, which is why so much scientific discovery is ultimately proven to be wrong. All reality exists in context, which means that reality is defined by far more variables than we can comprehend, much less measure and build into computations. The outer limit may well be infinity.As a result, any attempt to interpret reality and to use that “data” to think is reliant upon convention and limited representation. And convention is, by definition, imperfect.Pyrrho of Elis was a not-so-famous Greek philosopher who introduced what ultimately became known as the philosophical school of skepticism. It has been resiliently unpopular for reasons that psychologists can easily explain. Who wants to hang out with skeptics? And connection is ultimately what we all crave.Pyrrho’s skepticism related to dogma. A dogma is a rule or law or defined procedure or process. A convention is dogma as well. And Pyrrho’s issue with dogma was that whenever you lay it out you have opened the door for a duality—a truth and its exception. There are, quite literally, exceptions to every rule because reality is ultimately defined by an infinite number of variables that can’t be known by either a person or a machine. And that means that the exceptions can never be fully articulated no matter how much computational power you have at your disposal.Algorithms are ultimately nothing more than mathematical dogma. They can, by definition, never be complete. They will always be limited by probabilities, which is why they work at playing Go and translating language at a superficial level, but will never be the “black box” of human sentient consciousness that we all dream of. Never.I have witnessed the debilitating over-confidence in tech that Broussard speaks to repeatedly throughout my career in business. Technology is, in many ways, destroying modern business and, in particular, the social contract that employers used to recognize between employer, employee, and community.Business is consumed with reducing costs, which typically gets falsely interpreted as eliminating bodies. As a result, businesses typically want to automate everything, which, as Broussard explains, means that all of its processes need to be conceptually digitized. They must be restructured to accommodate the very real limitations of mathematics. Which ultimately means that they are often compromised and shaded by the very real personal biases of the person who made the digital conversion.That works some of the time in some circumstances. But not always. No company will ever be successful in fully automating processes like customer service, sales, quality, and innovation. To the extent they try, moreover, they risk disengaging the people they need to actually make those processes, perhaps assisted by technology, effective.Broussard has a strong political perspective. We all do. And there are portions of the book where she falls back on her investigative journalism and strays from the overall objective of the book. It’s always to make a relevant point, and she never quite abandons that objective, but the lapses are notable and just a tad distracting.In part, however, Broussard is admittedly trying to contrast the potential of tech socially and politically with the non-conformist, male-dominated libertarianism of the current tech industry. And that needs to change.The bigger irony, however, is that a culture and industry built on non-conformity is now ultimately turning back on itself and is not just promoting, but mandating, the ultimate in conformity. And the hidden risk is that unlike the conformity of things like organized religion, technology is forcing us to conform to norms and standards that we aren’t even aware of because the algorithms that drive our decisions and are filled with the human biases of the people who created them are largely hidden from view.And that is ultimately where, I suppose, algorithmic accountability reporting comes in. And I say, “bring it on.” It’s exactly what we need.In the meantime, however, we need to understand the overriding conceptual paradox of technology. Hal is a myth. The black box is a myth. The autonomous car is a myth (and if we give them broad access to our roads in your lifetime we will regret it). The potential of tech is not a myth. As long as we recognize that humanity is not obsolete, but it is biased.

I really enjoyed this book. It may not be for all readers. If you're not into computers and technology it might be a little boring for you. On the other hand, it might get your interested. The author is a very good writer with an almost conversational tone throughout the book that makes the reader feel like you are almost in the room with her having a pleasant chat. Definitely worth the read.

This was written pre-ChatGpt, but has a wide range of excellent concepts.

If you agree with the author 100% that all white males are chauvinists and over privileged then you will enjoy this book as it is well written and the author does a nice job mixing in personal anecdotes and clear explanations of the technology. If you want to read balanced facts and then reach your own conclusions, you are out of luck. This book is a missed opportunity to address an important issue. The author uses derogatory terms to describe people that might not share her views and discounts the efforts of others.

Loved this. Used it for a research paper on human computer interaction and it served me well!

This is one of the best written and informative explanations of digital technologies, especially artificial intelligence, that I've read. Broussard takes time to provide the context that is needed to understand how digital technologies work, including what they can and cannot do. Contrary to the claims of at least one reviewer, this is not a rant. Rather, it is a sane, articulate, thoughtful, evidence-based, well-informed, and thoroughly engaging explanation of digital technologies and their limitations, written by someone who is eminently qualfied through training and experience to write such a book. The dangers that Broussard warns about, which stem from "technochauvinism," the misplaced assumption that technoloigies are always needed, good, and effective, are very real. Those who embrace digital technologies uncritically will no doubt oppose this book. Keep this in mind when you read negative reviews.

The author made some valid points about the dangers and drawbacks of AI. However, I felt that 50% of the book (especially chapter 6) read like a SJW's rant. It is one thing to objectively analyze how politics affected computer science, the Internet and AI; it is quite another to insert disparaging remarks on gender-bias and racism on what felt like every other page starting from the second half of the book.Although it was a relatively short book (4-5 hours read), it left me feeling quite irritable. (Note: I do not work in tech/AI).

I think everyone with the slightest interest in AI should read this book. Skip the coding parts if you want. She has used it to illustrate how code works and how computers 'read' them, and hence how they work. It's clear that they only do as they are told to do, and since that telling is being done by us, humans, AI isn't going to best us anytime soon. She says probably never. Even self-driving cars is a very distant dream, which we may never achieve because the amount of data a human brain processes, is beyond the scope of even the most sophisticated AI system. I always knew this because it's basically common sense. Machines will only do what a human has programmed it to do. And even that, it will do with no filters and no application of mind, because it has NO mind of its own. Even the author agrees. AI will take over rote tasks but not ones that need ingenuity, imagination and creativity. Machines will only take over from the dullest, unproductive people on the planet. The rest of us with sparkly minds will continue to shine as always. Besides, the AI enabled gizmo can always be unplugged. So nothing to worry.

Die Authorin räumt mit vielen Praxisbeispielen mit dem Mythos der Künstlichen Intelligenz auf und zeigt deutlich das AI nun nichts mit Intelligenz zu tun hat wie sie bei vielen Menschen vorhanden ist. Sie räumt auch mit dem Mythos des autonomen Fahrens auf und erläutert warum dieses nicht möglich sein wird in einem Level 5 auf freien Straßen. Die Beispiele, wie IT, funktioniert sind auch für den Laien verständlich und der Experte darf weiter blättern.

Muito interessante a avaliação da IA feita pela Meredith...

Not well written Not worth the money

After a long "AI winter", it seems as if the pendulum has swung to the opposite pole. Hardly a day goes by without a news story or the other on the ground-breaking innovations made possible by artificial intelligence and its algorithmic cousins. Even in my own field, there are innumerable puff pieces (published in scientific journals, I am sad to say) about how "digital phenotyping", "ecological momentary assessments", "ChatGPT-based therapy bots" and the like could lead to paradigm shifts in mental health care.Yet is this enthusiasm justified? Can technology really play the role of a saviour, or is this just an example of us humans projecting our fantasies onto the latest technological advance? More importantly, are algorithms really the key to solving complex social problems, or will they simply worsen them - or even engender new ones?In this well-argued and lucid book, Meredith Broussard argues that we need to proceed with caution - and with a greater concern for ethics and human values. I, for one, found her arguments both convincing and valuable for future use in my own line of work.Highly recommended.