Book Review by Frank Cerwin:

Machines That Think:  How AI Works and What It Means For Us

By Inga Strümke, Reinwerk Publishing, 2026

I came across this recently published book at my local library.  Seems that I migrate to AI titles these days.  The title sounded like another primer on AI but I decided to pick it up anyway since it just required a library card to acquire a copy and I could return it tomorrow if disappointed.  However, I found that the title did not do justice to the content.  The author is a Norwegian physicist specializing in AI and Machine Learning and known in Europe for her work in AI ethics.  I did gain some new and valuable perspectives and insights from her book.

Here are just a few “food for thought” that I gleamed from the book:

• Neural Network models pick up whatever it believes as most useful in the data and doesn’t necessarily align with what we humans believe or know to be most important.
• When we use Machine Learning models to solve problems using data, we’re effectively assuming there’s a connection between the information in the data and the problem we want to solve.
• Machine Learning models have no way of knowing whether they have been given representative data – they simply optimize what’s given.
• Question if training data represents the world “we want to live in”, not the world as it exists or existed in the past.
• The golden rule of statistics is: “If you use one set of data to discover a phenomenon, you cannot use the same data to confirm the phenomenon, or any part of it.”

All of these perspectives are backed up with case studies and example by the author.  She describes how to collect data for different styles of Machine Learning.  Strümke puts an important focus on work being performed in Explainable AI (XAI) and the field of ‘Concept Detection’ to understand the specific conclusions created by AI.  I found the description of how deepfakes are created and how they improve themselves using two separate Neural Networks to be particularly eye-opening as well as scary.

The final chapters of the book address ethics related to what AI means to society, employment, and job loss.  The author states that “Ethics is not about finding ‘the right thing to do’”.  But rather, “It’s about weighing different interests against one another”.  An interesting definition that she substantiates with her observations and experience.  Further, Strümke provides different approaches to managing risks of unethical behavior with a set of ethical building blocks.  Lastly, if you’re concerned about data security and how your data may not be as safely protected as you think it is, you will learn how “Membership Attacks”, “Jailbreaking”, and “Adversarial Images” can expose personal information and fool AI.

In conclusion, ‘Machines That Think’ is a worthy read for every data management professional.  It has an overall emphasis on data and provides important insights and approaches that we all should know as we continue a journey into AI and Machine Learning data processing.