Book Review: Machines That Think by Inga Strümke

Machines That Think stands out as a well-structured and thoughtful introduction to artificial intelligence, balancing technical clarity with deeper philosophical inquiry. Rather than rushing into modern buzzwords, Inga Strümke takes a deliberate approach, guiding readers from the earliest foundations of computation to the complex ethical and societal challenges shaping AI today.

Building from First Principles

The book opens with a grounded exploration of early computational ideas, beginning with algorithms and analog computers. This historical framing is not just background filler—it establishes the logic behind how machines process information and why AI evolved the way it did. By walking through these origins, Strümke ensures that readers understand AI as a continuum rather than a sudden technological leap.

One of the more valuable early distinctions she introduces is the difference between AI and cybernetics. While both fields deal with systems, feedback, and control, cybernetics focuses more on regulation and communication within systems, whereas AI is concerned with creating systems that can perform tasks associated with intelligence. This distinction helps clarify why modern AI developed along its current trajectory, rather than following the more biologically inspired path of early cybernetics research.

From Brute Force to Learning Systems

The evolution of AI is perhaps most clearly illustrated through its relationship with games, particularly chess. The contrast between Deep Blue and AlphaZero is used effectively to demonstrate how dramatically AI methodologies have shifted.

Deep Blue relied heavily on brute-force computation, evaluating millions of possible positions per second to determine optimal moves. It was powerful, but fundamentally limited by its reliance on pre-programmed evaluation functions and raw processing power. In contrast, AlphaZero represents a paradigm shift. Instead of being explicitly programmed with strategies, it learns by playing against itself, discovering patterns and strategies that even human grandmasters had not previously considered.

This transition highlights one of the core themes of the book: modern AI is less about explicitly telling machines what to do and more about enabling them to learn from data and experience. It also underscores a deeper philosophical point—AI systems are no longer just tools executing instructions, but systems capable of generating novel solutions.

Symbolic vs. Non-Symbolic AI

A particularly strong section of the book focuses on the distinction between symbolic and non-symbolic AI. Symbolic AI, which dominated early research, relies on explicit rules and logical representations of knowledge. It is structured, interpretable, and deterministic, but often struggles with ambiguity and real-world complexity.

Non-symbolic AI, particularly neural networks, operates very differently. Instead of rules, these systems learn patternsectly from data. This makes them far more adaptable and powerful in tasks like image recognition, natural language processing, and pattern detection. However, this power comes at the cost of interpretability. Neural networks often function as “black boxes,” making it difficult to understand how they arrive at specific decisions.

Strümke does an excellent job explaining this trade-off without oversimplifying it. She shows that the shift toward non-symbolic AI is not just a technical evolution, but a fundamental change in how we approach intelligence itself.

Understanding Neural Networks and Perception

The sections on image classification and neural networks are among the most engaging in the book. Strümke explains how neural networks process visual information in layers, gradually transforming raw pixel data into higher-level abstractions. Early layers detect simple features such as edges and shapes, while deeper layers identify complex patterns like objects and faces.

This layered approach allows machines to perform tasks that once seemed uniquely human. However, the book makes it clear that this “understanding” is not the same as human comprehension. AI systems recognize patterns statistically, not semantically. They do not “know” what an object is in the way humans do—they identify it based on learned correlations.

At the same time, there are domains where AI systems surpass human capabilities. In image recognition, for example, machines can detect subtle patterns and anomalies that are invisible to the human eye. This creates an interesting paradox: AI systems can outperform humans in specific tasks while still lacking true understanding.

Explainability, Ethics, and Regulation

As the book progresses, Strümke’s voice becomes more pronounced, particularly when discussing explainable AI and data privacy. She highlights a critical issue in modern AI: as systems become more powerful, they also become less transparent.

Explainable AI aims to address this by making machine decisions more interpretable. This is especially important in high-stakes domains such as healthcare, finance, and law, where understanding the reasoning behind a decision is crucial. Strümke emphasizes that without explainability, trust in AI systems becomes difficult to establish.

Her critique of data privacy laws is particularly thought-provoking. She argues that many existing regulations are both restrictive and ineffective. On one hand, they can slow down innovation by limiting access to data. On the other hand, they often fail to adequately protect users due to loopholes and enforcement challenges. This dual failure highlights the complexity of governing rapidly evolving technologies.

Creativity and the Limits of AI

Another compelling theme in the book is the question of creativity. With the rise of AI-generated art, music, and writing, the line between human and machine creativity is becoming increasingly blurred.

Strümke approaches this topic with nuance, asking whether AI systems are truly creative or simply recombining existing human-generated content. While AI can produce impressive outputs, its creativity is fundamentally different from human creativity, which is shaped by consciousness, experience, and intention.

This discussion extends into broader ethical considerations, including ownership, originality, and the value of human input. Rather than offering definitive answers, the book encourages readers to engage with these questions and consider their implications.

The Collective Action Problem in AI

One of the more subtle but important ideas explored is the collective action problem in AI deployment. As AI systems become more powerful, the decisions of individual organizations can have wide-reaching consequences. However, coordinating action across governments, companies, and institutions is inherently difficult.

This creates a situation where short-term incentives can conflict with long-term societal interests. Strümke highlights the importance of collaboration and governance, while acknowledging the challenges involved in achieving it.

Looking Toward AGI and Beyond

The book concludes with a forward-looking discussion on Artificial General Intelligence (AGI) and superintelligence. Rather than indulging in speculation, Strümke grounds her analysis in current technological trends and limitations.

She explores what it might take to achieve AGI, including advances in computation, data, and algorithm design. At the same time, she considers how society might adapt to increasingly capable AI systems. This includes not only technological changes, but also shifts in education, policy, and economic structures.

The ending feels appropriately balanced—optimistic about the potential of AI, but realistic about the challenges ahead.

Final Verdict

Machines That Think is a well-rounded and engaging addition to the growing body of literature on artificial intelligence. It succeeds in making complex topics accessible without sacrificing depth, while also addressing the ethical and societal implications of the technology.

For readers looking to build a strong foundation in AI—while also exploring its broader impact—this book offers a thoughtful and comprehensive perspective. It is both an introduction and an invitation to think more critically about the role AI will play in shaping the future.