The Myth of Artificial Intelligence by Erik J. Larson
The common assumption is that we are on an inevitable path toward creating human-level artificial general intelligence (AGI). Erik J. Larson argues this is a myth. He makes a compelling case that the current methods of AI, which rely on inductive reasoning
by crunching massive datasets, are fundamentally different from human thought. Humans use a form of reasoning called abductive inference—essentially educated guesswork and context-based intuition—which we have no idea how to program. Larson challenges the hype by showing that we aren't just a few breakthroughs away from AGI; we may not even be on the right path at all. The book suggests our cultural obsession with superintelligence is distracting us from the real-world limitations and capabilities of AI today.
Weapons of Math Destruction by Cathy O'Neil
Many assume that algorithms, being based on math, are inherently neutral and objective. Mathematician Cathy O'Neil dismantles this idea, revealing how many of the algorithms governing our lives are what she calls "Weapons of Math Destruction" (WMDs). These models are often opaque, unregulated, and operate at a massive scale. Instead of eliminating bias, they can amplify and codify it. O'Neil provides real-world examples, from how teacher performance is graded to how predictive policing targets certain neighbourhoods, showing that these algorithms often punish the poor and reinforce existing social inequalities. The book is a powerful argument that fairness doesn't automatically emerge from data; it must be consciously designed into our systems.
Atlas of AI by Kate Crawford
AI is often thought of as something ethereal and abstract—code living in "the cloud." Kate Crawford challenges this by grounding AI in its physical reality. Her book reveals the immense, and often hidden, planetary costs of building artificial intelligence. This includes the carbon footprint of massive data centers, the destructive environmental impact of mining for the minerals needed for our devices, and the exploitation of low-wage human labor required to label data and keep the systems running. Crawford's work forces us to see AI not as a disembodied intelligence but as a material industry built on natural resources and human effort, with profound political and environmental consequences.
The Age of Surveillance Capitalism by Shoshana Zuboff
It is easy to assume that the data collected about us by tech companies is primarily used to improve their services. Shoshana Zuboff argues this is a dangerous misunderstanding of the current economic model. She posits that a new form of capitalism, "surveillance capitalism," has emerged. In this system, our personal experiences are claimed as free raw material, translated into behavioural data, and then used to create "prediction products." These products are sold to other businesses that want to influence and control our future behaviour. Zuboff's thesis challenges the idea that AI is just a tool; she frames it as the engine of an economic system that threatens human autonomy and democratic norms by treating human life as a resource to be mined and manipulated for profit.
The Alignment Problem by Brian Christian
A common assumption is that once AI becomes powerful enough, we can simply instruct it to follow human values. Brian Christian’s book explores why this is one of the hardest problems in computer science. The "alignment problem" is the challenge of ensuring that an AI's goals are truly aligned with our own. Christian details the many ways this can go wrong, from biased training data creating discriminatory outcomes to AI systems finding clever but destructive loopholes to achieve the goals we set for them. The book dives into the complex, multidisciplinary effort to encode nuanced human concepts like fairness and safety into machine learning systems, revealing that teaching a machine our values is far more difficult than teaching it to play a game.
