What Is Intelligence?: Lessons from AI About Evolution, Computing, and Minds

The chapter on hydrothermal vents provides an incredible foundation for Aguera y Arcas’s central argument that intelligence isn't some magical spark but a byproduct of persistent chemical cycles. It’s a dense read, yet the prose remains remarkably clear as it connects the dots between early metabolic systems and the transformer architectures behind modern LLMs. By framing AI as a natural extension of biological evolution, the author strips away the sensationalism often found in current tech writing. I appreciated how he explains that systems surviving in hostile environments are essentially "predicting" their surroundings through adaptation. This isn't just about code; it’s about the thermodynamics of life itself. Even if you aren't a computer scientist, the connections drawn between bacterial behavior and neural networks are absolutely mind-expanding.

As someone who has spent years tracking the development of neural networks, I found this to be one of the most coherent explanations of why self-supervised learning changed everything. The author argues that moving away from rigid, label-based training was the key to unlocking general-purpose reasoning, and the logic here is ironclad. He weaves together concepts like fractals and internal models to show how "minds" are often composed of smaller "subminds" in a constant feedback loop. It is rare to find a tech leader who can write with such philosophical depth without losing the thread of the actual science. This isn't just a book about AI; it's an exploration of what it means to be a processing entity in an entropic universe.

Picked this up after seeing the online version and honestly, it’s a masterpiece of cross-disciplinary synthesis. The way the author connects the history of computing—from Leibniz and Turing—to the way modern LLMs predict the next token is nothing short of brilliant. It’s a grounded perspective that doesn’t treat AI as a sudden alien intrusion but as a continuation of a process that started billions of years ago. I particularly enjoyed the discussion on collective intelligence and how humans are becoming part of a larger hybrid system. If you want to move beyond the shallow "AI alarmism" and actually understand the underlying mechanics of intelligence, this is the definitive text for the current era.

This is exactly the kind of deep-dive we need in an era of superficial AI hot takes. Instead of focusing on the "what" of AI, Blaise explores the "why" by looking at the very origins of chemical systems on Earth. The connection between self-replication and internal modeling is explained so clearly that it feels obvious in hindsight. I was particularly fascinated by the idea that language is a compressed representation of thought that allows us to share our internal models with others. It’s a beautiful, hopeful book that suggests our future with AI will be one of shared, collective growth rather than competition. This is essential reading for anyone who wants to understand the long-term trajectory of our species and our tools.

This book is a refreshing departure from the usual "AI will save or kill us all" tropes that dominate the shelves lately. Truth is, I wasn’t expecting a Google CTO to dive so deep into the chaotic flight patterns of moths just to explain predator-prey dynamics and strategic randomness. The core idea—that intelligence is basically just a sophisticated internal model for predicting the future—is handled with a lot of nuance and historical context. My only real gripe is that some of the sections on cybernetics felt a bit like a history lecture, which might slow down readers looking for a more futuristic focus. Still, the way it bridges the gap between biology and computation is incredibly satisfying. It’s a solid 4-star read for anyone tired of the hype cycle.

Ever wonder why a tiny bacterium can navigate toward nutrients while your smart home device still struggles with basic context? Blaise Aguera y Arcas tackles this by showing that even the simplest life forms are performing a version of statistical inference to survive. I loved the parts about "theory of mind" and how it underwrites our ability to make purposive decisions beyond mere autopilot. The writing style is didactic but stays engaging enough to prevent it from feeling like a dry textbook. My one complaint is that the physical book lacks some of the cool animations found in the online version, which makes certain concepts like fractals a bit harder to visualize. Definitely a book that requires a second reading to fully absorb.

Look, if you’re looking for a quick beach read about robots, this isn't it, but it’s probably the most important book you’ll read on technology this year. The author’s main contention is that intelligence emerges whenever a system learns to model itself and its world to overcome entropy. This "backwards causality" of anticipating the future is a fascinating way to look at everything from a moth’s flight to a chatbot's response. Personally, I found the middle chapters on symbolic logic a bit dry compared to the fascinating stuff on biological cells. However, the overarching argument that intelligence exists on a broad spectrum rather than being a human-only trait is incredibly persuasive. It’s a dense, rewarding challenge that shifts your worldview.

After hearing about this on a few tech podcasts, I decided to see if the hype was real. The book is a sweeping reinterpretation of what "thinking" actually is, moving away from human exceptionalism toward a more universal definition. Aguera y Arcas explains that prediction is the heartbeat of intelligence, whether it's happening in a brain or a silicon chip. The prose is sophisticated, using long, complex sentences to build a case before dropping a punchy, clarifying insight. While I struggled with some of the more intense thermodynamics sections, the payoff was a much clearer understanding of how life and computation are two sides of the same coin. It really makes you rethink your own place in the evolutionary timeline.

Not what I expected given the pedigree of the author. While Aguera y Arcas is clearly brilliant and offers a unique "insider" view from Google, the book often feels like it's trying to cover too much ground at once. We jump from the origins of life to thermodynamics to large language models, and the thread occasionally gets lost in the weeds of technical definitions. To be fair, the section on theory of mind was provocative, but I found the dismissal of the "hard problem" of consciousness a bit too hand-wavy for my taste. It’s a very academic-leaning text that requires a lot of patience. If you’re already well-versed in information theory or neurobiology, you might find about 70% of this redundant.

Frankly, I found the author's optimism regarding the "collective intelligence" of AI-human systems to be a bit too convenient for someone in his position at Google. The book is undeniably smart and well-researched, but it often brushes over the ethical and social vulnerabilities of these complex systems in favor of grand evolutionary theories. I enjoyed the sections on split-brain patients and how they relate to internal models, but the transition into large language models felt like it was missing a few steps of critical analysis. It’s a very "insider" view that sometimes ignores the darker implications of the tech we are building. It’s worth reading for the science, but keep a skeptical eye on the final predictions for our shared future.