BOOK REVIEW: Artificial Unintelligence How Computers Misunderstand the World by Meredith Broussard
By Miras Tolepbergen
Since the first computer launched in 1946, it only took half a century for humans to figure out how to integrate digital technology and society. Today, almost every field has experienced a computational turn, including the humanities, social sciences and natural sciences. It is not only the scientific fields that are infected: visual artists create multimedia art and sculptors use 3D printers. People have also begun to use digital technology in harmful ways, including the dissemination of “fake news”, illegal surveillance, the construction of armed drones and the selling personal data, to name a few. This is certainly not the better world that early technology geniuses promised. Despite the current highly-computerized lifestyles, human nature has not changed; we live in the same world with the same types of social issues that have always been around.
In Artificial Unintelligence: How Computers Misunderstand the World (2018), Meredith Broussard illustrates how technology cannot achieve all things many people imagined it could do because there are jobs and tasks for which machines cannot replace humans. Technological innovation cannot solve the essential problems of human nature. Why then, the author asks, do people persist in believing in an imminent technological utopia? This book illustrates the limits of what humans can, and should do with technology. The author argues that understanding these limits will help humans make better choices and have conversations about the ways technology can be used to make the world a better place.
The author is concerned with the assumption of “techno-chauvinism,” which is the belief that computers are objective or unbiased because they are based on mathematical principles, and thus have the potential to solve all social problems and create a digitally enabled utopia. Broussard identifies two major problems with this assumption. First, mathematical logic, based on computers, does not always translate well into the offline world. In other words, computers manipulate symbols without referring to their interpretations, while humans can interpret socially constructed systems in meaningful ways. Second, technology geniuses, who rely on those symbols may lack an understanding of cultural systems built around human beings. Thus, technological innovations cannot be flawless, which undermines techno-chauvinists’ imagined perfect world. Therefore, Broussard argues, people should be careful with how technology evolves. Biases, both in computational systems, as well as in the technology industry should be reduced. It should be ensured that technological innovations are designed to follow the rule of law and respect social norms.
The author argues that human assistance systems, not human replacement systems should be the focus of the technology industry. For instance, the ongoing race to build self-driving cars illustrates the fundamental limits of computing. The core problem is that machines lack sentience, which is the cognitive capacity for reasoning about the future. Driving involves performing many different tasks simultaneously, and autonomous technology that we have today is incapable of achieving that. Thus, a self-driving car would not be able to overcome obstacles the way that a human would. In other words, machine learning (ML) is great for routine tasks inside a fixed universe of symbols, not in the complex real world where things could change constantly and unpredictably.
Diversity is the key to creating a more technologically just world. For instance, social media, where popular photos collect many “likes”, creates the assumption that popular means good. The author claims that this type of mistake is quite common among computational researchers who may lack critical reflection on social values and human behavior. The problem is that algorithms are affected by unconscious biases because people design them. Since it is rarely intentional, we should be critical about the things we know can go wrong. The author claims that individuals must investigate the wider applications and implications of their technical choices, and be prepared for the fact that they might not like what they find. If a machine simply replicates the world as it is now, there is no hope for a more just society.
Broussard also explores the difficulty of talking about technology, which tends to lead to many misunderstandings. The difference between mathematical language, which is highly precise, and everyday language, which is arbitrary, creates a communication problem that exists in computational culture today. For instance, ML implies that a computer is sentient and that it “learns”, although “learning” is a term usually used to refer to human beings and occasionally, animals. However, for computer scientists, learning means that the machine can become better at its programmed, routine and automated tasks. This understanding is very different from “acquiring knowledge by machine”. “Learning” by the machine does not imply “intelligence”. According to the author, computer scientists have redefined the term “learning” so that it refers to their work.
Moreover, Broussard demonstrates that most of the ideas about digital technology stem from a single small group of elites. This group historically has lacked diversity, and their beliefs are embedded in the design of a technology they create. For instance, most of the tech geniuses’ anti-social behavior has been ignored as long as they have continued to innovate, which shows the prioritization of an efficient code above human interaction. The author invites the reader to rethink and revise such beliefs, because understanding how computers work is not the same as understanding how society works. Thus, the author adds, people should not aspire to be governed by computational systems designed by individuals who lack an understanding of cultural systems built around human beings.
Meredith Broussard’s critique is reasonable, particularly regarding so-called autonomous technology. Pure mathematical logic and the absence of sentience are the strengths of machines. Meanwhile, human beings are irrational, and sentience could be the main reason for that. Thus, sometimes individuals need thinking models based on pure calculations to make better choices, while autonomous technology needs human involvement in the decision-making loop regarding ethics and social norms. Thus, as Broussard states, technology should assist us, not live our lives. In other words, a combination of tools based on pure mathematical logic and the human ability to express sentience could help us lead better lives.
The primary strength of the Artificial Unintelligence: How Computers Misunderstand the World is that it is written majorly for non-expert audience and the message it conveys is clear: studying of programming should be realized in tight collaboration with social sciences. In other words, relying solely on hard-sciences might lead to chaotic and unintended consequences due to the limitations of artificial intelligence. Thus, this book is for anyone interested in technological innovation. A key weakness is that the author only designates the limitations of artificial intelligence and broadly mentions in which direction its development should head, but provides no concrete solutions to the discussed issues. Though the book is a good introductory material into the artificial intelligence world, it will not be enough for those who wish to dig deeper into the field.
Miras Tolepbergen received his BA in Politics and International Relations from Nazarbayev University in Kazakhstan. Currently, he is an MA student in International Relations and Diplomacy program at Shanghai University, PRC. His interests include IR Theory; Middle Eastern Politics; and the Sociology of Religion, Science, and Technology.
Image: Cover of Artificial Unintelligence: How Computers Misunderstand the World
