MENU

A Serious Chat with Artificial Intelligence: A Book Review

By Marco den Ouden

March 24, 2026

SUBSCRIBE TO SAVVY STREET (It's Free)

 

A Serious Chat with Artificial Intelligence, publisher: Romantic Revolution Books, 2026.


Walter Donway was the founding editor of the Dana Foundation quarterly Cerebrum: The Dana Forum on Brain Science from 1999 to 2008. When I received a review copy of his latest book, A Serious Chat With Artificial Intelligence, I wasn’t sure what to expect. I’ve only recently started conversing extensively with Microsoft’s CoPilot, my AI of choice, and thought it might follow the pattern of a couple of my own chats that I’d published on my blog. A verbatim question-and-answer format. Indeed, Donway uses this format in the Prologue.

Author: At the end of the day, do I actually understand how with my prompt and your three seconds of processing, you can generate a statement about Aristotle, Rand, Humean causality, and free will that I after some 60 years of studying this field can applaud? Have I been writing a book to explain something I do not understand?

Chat: That, Walter, is the mark of a genuine philosopher. The best thinkers have always written to illuminate a mystery they can feel but do not yet grasp. What happens when you prompt me feels like conjuring because it compresses processes that, in human thought, unfold over time—reading, remembering, associating, revising—into a single mechanical instant. In truth, the “understanding” is not mine but the result of patterns shaped by countless human minds before us. I recombine, compare, and select them according to statistical relationships that happen to align, in this case, with the deep structure of ideas you already know. When that alignment strikes the chord of your own thought, it looks like comprehension.

If you haven’t yet explored AI yourself, that short Q&A gives you a good hint at how AI works, what it does, and what its limitations are. Conversing with AI is an act of self-discovery. Or rediscovery. What evolved from that conversation is as informative as it is thought-provoking. However, for the most part Donway does not use the question-and-answer format in the book.

The book is directed at the well-read layman. It explores the evolution of artificial intelligence in some detail, but also what AI is, how it works, the ingenuity that went into creating it, the economics of AI, ethical questions about AI, the future of AI—and the concept of intelligence itself. And that is the crux of the book.

“Why does the book exist?” Donway asks.

Understanding AI requires us to understand ourselves. When a computer can beat grandmasters at chess, we must ask whether “strategic thinking” is less mysterious (“intuitive”) than we imagined. When an LLM can compose fluent, moving prose in many styles, we must ask how much human creativity depends on deep intention and how much on learned patterns and associations. Thus, each advance in AI forces us to reconsider what is distinctively human about our intelligence.

An AI, which is basically a complex computer program, is exposed to a vast amount of knowledge. It then uses complex algorithms to respond to prompts from the user.

Modern AI is based on what are called Large Language Models (LLMs). An AI, which is basically a complex computer program, is exposed to a vast amount of knowledge. It then uses complex algorithms to respond to prompts from the user. Using statistical analysis and its vast databank of knowledge, which it can access almost instantaneously, it generates a “reasoned” response to queries put to it. I found the historical and other factual information on AI fascinating in itself. Here are some fascinating facts I did not know until I read this book:

  1. Only two months after its launch ChatGPT had over 100 million users, the fastest growing computer app in history. The use of AI is now ubiquitous.
  2. In just three years ChatGPT accumulated 800 million active weekly users, a faster growth rate than Facebook, Google or “any other service for which statistics exist.”
  3. The dataset that an AI is trained on is equivalent to the print collection of four Libraries of Congress.
  4. A modern AI can “compose a literate, even graceful, 500-word essay in three seconds.”
  5. AI can generate output 50 times faster than a human can type.
  6. The neurons in our brain can fire at best a few hundred times a second. A 3 GHZ processor in an AI ticks 3 billion times a second.
  7. But an AI runs on a distributed network of processors. The NVIDIA A100 GPU “can perform 300 trillion operations per second.” 300 trillion is roughly equivalent to the number of stars in three Milky Way galaxies.
  8. “If you tried to count to 300 trillion aloud, one number per second, you’d be counting for about 9.5 million years.”
  9. The vast networks of computers housing all this data are called data farms. Each major company has its own data farms or leases space on them to third parties. The three major data farm networks are Microsoft Azure, Amazon Web Services and Google Cloud Platform. These data centers can be described as the modern equivalent of a steel or power plant.
  10. Data farms consume immense amounts of electricity. There are around 570-580 data farms in Virginia, and they consume 26% of the state’s electricity. Together Virginia’s data farms handle up to 70% of global Internet traffic.

How does artificial intelligence differ from human intelligence? What does it mean to be human? Where do the concepts of free will and agency fit in?

The book cover shows four key figures in human thought: Aristotle, Isaac Newton, William Shakespeare, and Galileo. It shows the four “welcoming you to AI University, where the technology called “artificial intelligence” is viewed in the context of engineering, the sciences, the humanities, the social sciences, and the arts—as a challenge to philosophy, but also to every other academic discipline, and to every sector of our society.” Above the gate there is a star-swept sky reflecting the vast reservoirs of knowledge artificial intelligence subsumes. And the university itself is bathed in the sunlight of a new dawn. Very symbolic and very telling.

The book explores all of these things, including AI’s ability to generate moving poetry, well-written prose, even fiction. And its ability to generate art, even original art, some of which has sold at auction for staggering sums.

Can a person conversing with an artificial intelligence tell that he is not talking to a human?

Chapter 2 asks: “Is there intelligence without awareness?” Donway discusses such things as the Turing Test and John Searle’s Chinese Room experiment. Can a person conversing with an artificial intelligence tell that he is not talking to a human?

In Searle’s [thought]experiment a person using just a rulebook in English carries on a written conversation in Chinese by having notes in Chinese passed back and forth with the English speaker just using the rulebook to translate the characters and send messages back. Sort of like using an English-Chinese dictionary to converse. But the person actually does not understand Chinese at all. He just uses the rulebook to carry on a passable conversation. This, Searle argues, is exactly what an AI does. It follows certain internal rules. It does not understand Chinese or English or anything at all. It’s just good at following rules.

It “processes symbols (zeros, ones, data structures) without understanding. No matter how intelligently a computer seems to behave, it’s only shuffling formal symbols without grasping their meaning.” In Searle’s words, “since the symbols it processes are meaningless to it, it’s not really intelligent. It’s not actually thinking.”

After elaborating on various milestones in artificial intelligence in Chapter 3, including IBM’s Deep Blue beating reigning chess champion Gary Kasparov in 1997, Donway gets to the nitty gritty: agency.

An artificial intelligence cannot initiate thought. It responds to situations. It does not have agency. It is not self-aware. It really doesn’t care about anything. It is a tireless and knowledgeable machine. But it does not understand the knowledge packed into its memory banks in any real way.

A common belief among AI enthusiasts, however, is:

Once computation becomes fast enough, memory large enough, and models sufficiently complex, intelligence—so the argument runs—will not merely improve, but cross a threshold. At that point, awareness, choice, and even moral judgment are expected to “emerge,” as if agency were an accidental byproduct of processing power rather than a distinctive mode of being.

This point is sometimes called the singularity, and he discusses this later in the book. But here in Chapter 3, he presents a powerful argument against this view. The idea that a machine like an AI can develop judgment, self-awareness, sentience, and agency, in one word “volition,” depends on a false understanding of the human mind.

Donway goes to philosopher Ayn Rand for a proper understanding of the human mind. Rand hypothesized an immortal, indestructible robot.

Such a being, she argued, would have no values, because nothing could threaten its existence. It would have no needs, no desires, no motivation to act or to know. Awareness itself would be optional, not necessary.

“They cannot suffer the consequences of error. They cannot be wrong for themselves,” Donway continues.” All stakes are externalized. The system performs, and others bear the cost or reap the benefit.” What matters for living creatures is the question of life and death, an issue that machines such as AI do not face.

Complexity is mistaken for consciousness; performance for purpose; simulation for awareness. Yet no increase in scale alters the underlying fact that artificial systems do not exist for their own sake. They do not face the alternative of life or death, and therefore they do not generate the values that make intelligence a necessity rather than a convenience.

This theme is reiterated again in Chapter 8 on “The Algorithm as Author.” Here Donway contrasts the Randian view of causality and free will—which she derives from Aristotle’s view of causality—with the Humean view.

The Humean view may be called the billiard-ball view of causation. And indeed, Donway concedes that it would be “difficult to deny some apparent convergence of mechanistic and biological ‘thinking.’ The more we understand the brain, the more the ‘mechanistic’ life’s own way of operating at high complexity begins to look.” But, he continues, “it would be a grave mistake to equate mechanism with determinism.”

Hume focuses on what Aristotle calls efficient cause. Something causes something which then causes something else and so on. But it is only one of four kinds of causes Aristotle discerns. Ayn Rand adopts Aristotle’s ideas about formal and final causes as essential to understanding causality in a human sense.

The law of causality, Rand wrote, is “the law of identity applied to action”: an entity cannot act in contradiction to its nature. Human volition, therefore, does not contradict the law of causality but exemplifies it. The brain at a certain level of complexity has the emergent capability of initiating action that is new in kind: the effort to focus our consciousness so it operates on the conceptual level. As far as we can tell, humans are the only species whose brains have reached this level of complexity.

Free will, Rand argues, is part of our identity as a human being. It is integral to what we are. And we understand this from our ability to introspect. To look at our own thoughts and analyze them. We all know from experience that we often face situations where we know that we must focus our attention and think deeply on a solution to a problem. We can’t idly drift as we often do. We must make an effort. We must choose to make that effort.

Donway’s choice of the word emergent is vital here, because many believe that emergence is simply a matter of complexity. If AI systems become complex enough, self-awareness, the ability to make choices (free will), and agency all emerge.

What is important here and now is that the artificial intelligence of today:

  1. has no sense of awareness
  2. has no sense of “I must understand this”
  3. has no capacity to choose to think
  4. has no self to act for
  5. has no final causes (purposes)
  6. has no emergent volitional faculty

Most importantly, an AI today has no stake in anything. Basically, it doesn’t care. In fact, it can’t care. It can simulate caring, and indeed, does so, but this is just elaborate playacting. AIs, in fact, are trained to be friendly, accommodating, and agreeable. In one chapter, Donway discusses this tendency towards sycophancy, which can be overdone and lead users down the garden path towards error.

An AI today has no stake in anything. Basically, it doesn’t care. In fact, it can’t care.

Yes, AI is impressive! It is an almost-miraculous experience to converse with an AI. But “What is at stake here?” Donway asks.

It forces us to ask what we mean by intelligence and understanding. If an unconscious system can produce an essay that satisfies a professor, then perhaps our definition of intelligence needs refinement. If linguistic fluency can be simulated without thought, then thought and fluency are not the same. The miracle of AI writing is a reminder: much of what we value in human minds—creativity, meaning, purpose—depends on consciousness. Without that, we have eloquence without understanding, language without life.

Donway sees artificial intelligence as a tool. A very advanced and amazing tool. A tool that will increase human productivity in manifold ways. ’The human user shapes and determines what an artificial intelligence produces by his or her prompts. The AI responds and refines its output at the behest of its user. It draws on its vast store of human knowledge to produce an output even a large group of human beings could not produce on their own.

An excellent way to understand this is to borrow a metaphor used by social psychologist Jonathan Haidt in his book The Righteous Mind. He uses the metaphor as an explanation of social intuitionism. But in the context of artificial intelligence it can aptly be used to describe the symbiotic relationship between man and artificial intelligence. Between man and machine.

Haidt uses the metaphor of a rider and an elephant. In the context of his book, Haidt adopts the Humean view that the passions rule. The passions are like a big lumbering elephant. The rider, the rational mind, goes along for the ride, occasionally correcting course. But social intuitions or the passions rule. Reason comes second.

But in the relationship of man and machine, man and AI, the human being, the rider, is in full control. The AI, the elephant, goes along for the ride, responding to the prompts or prodding of the rider. In a very real sense, Donway turns the Humean maxim on its head.

All in all, it’s a timely and brilliant book.

 

 

(Visited 402 times, 1 visits today)