Home Robotics R. Kurzweil – How to create a mind (“How to create a mind”). A brief overview of the book

R. Kurzweil – How to create a mind (“How to create a mind”). A brief overview of the book

by admin

R. Kurzweil – How to create a mind ("How to create a mind”). A brief overview of the book
Ray Kurzweil is quite an interesting person. To begin with, he was the first person to invent music synthesizers in 1965. He won a whole series of awards for his inventions in various fields and published several books (The age of intelligent machines – 1990, The 10% solution for a healthy life – 1993, the age of spiritual machines – 1998, The singularity is near – 2005), made a film and even founded (together with Google and NASA) Singularity University. So when word came out that he was releasing a new book on how to create a brain (How to create a mind – the secrets of human thought revealed) – I immediately pre-ordered it.
And for those who doubt whether to buy it, and for all those interested, I offer a kind of review: but what exactly did the author want to say?
There are 11 chapters + an epilogue in just 321 pages, so for those who are familiar with English (and Ray loves and knows English 🙂 ), getting to know the book won’t take long, quite possible to master in a weekend.
When I pre-ordered, I was pretty sure I was going to get a manual that detailed how to actually make the mind and take over the world. Reality, as usual, made its own adjustments. It turned out that the book is basically an excursion into the history of philosophy in general and computer science in particular. The actual methods used by Ray in his projects (such as, for example, the speech recognition project) are given only one chapter. Before that is mostly the story of how these methods came to be, and after that, why these particular methods should be taken as a recipe for mind creation.
So, let’s go through the chapters : the introduction and the first two chapters ("Thought Experiments on the World" and "Thought Experiments on Thinking"), in the best tradition of "starting from afar", tell how human thought evolved and what inventions were made at the tip of the pen. History buffs will find it interesting; for everyone else, it’s educational. It is true that already in the second chapter he focuses on the fundamental points on which the further theory of "thought-making" will be based:
– "…our memories are consistent in nature. They can be reproduced in the same order as remembered. We cannot directly access memories in an arbitrary order."
– "There are no pictures, videos or sounds in the brain. All of our memories are stored as sequences of patterns (here I don’t even know what word would properly convey the meaning). Unused memories fade over time."
– "We can recognize an object even in part (sight, hearing, touch) and even if the object has undergone significant change. Our perception has the ability to distinguish characteristics that are unaffected by changes in the real world."
– "Our conscious perception actually depends on how we think in the moment. That is, our brain is constantly predicting what should happen in the next moment and how we should feel. This anticipation is what influences our perception."
– "Our memory is not a list of a thousand remembered actions, but rather each of our remembered actions is an elaborate hierarchy of nested actions. This same hierarchy is also evident in our ability to recognize objects and situations."
Reading this, I had a feeling I’d seen it somewhere before. Well, yes, of course – Jeff Hawkins, On Intelligence. I highly recommend it, by the way. Actually it is a set of requirements put together which inventors of artificial intelligence are trying to implement in their creations.
And here we come smoothly to the third chapter: the neocortex model, a theory of thinking based on pattern recognition. The chapter begins with a rather interesting estimation of the neocortex (cortex) capabilities based on the number of neurons, neural columns and 100’000 "pieces of knowledge". I don’t know where he got this information, but the calculation is roughly as follows: Kasparov learned 100’000 chess positions, Shakespeare composed using 100’000 language constructions, the typical medical professional knows about 100’000 concepts, and considering that many concepts are stored in excess (100 to 1, as he claims), the human brain can learn 10 million patterns. But by a simple assertion he brings this number to 300 million as well. This coincides with the number of "elementary recognizers", which he estimates to consist of about 100 neurons each (the neocortex has 30 billion neurons: 500 000 cortical columns with 60’000 neurons in each). He gives these estimates without references, so it looks very far-fetched, but beautiful. It is clear that the elementary recognizer is a perceptron with weight coefficients, and everything is organized in a hierarchical neural network. And the whole chapter he discusses the role of patterns and their recognizers, how they relate to each other and influence each other, ending with a mention of his development of a voice recognizer based on hierarchical hidden Markov models: the approach seems to work.
Briefly the whole chapter can be described as follows : patterns -> elementary recognizers -> neocortex -> Profit.
The fourth chapter has already begun to bring surprises. In describing the biology of the neocortex, he refers to the Blue Brain Project and to the words of its creator, Henry Markram, about the groups of neurons found in which synaptic connections and their properties are well predictable and non-random. Markram himself assigns them the role of elements of innate memory, and Kurzweil – the role of the very "elementary recognizers" (there are about 100 neurons in these groups). Further in this chapter, we discuss the 3-dimensional data backbone running through the entire neocortex discovered in March of this year, the neuroplasticity provided by the single method of signal processing in the entire neocortex, and the possible role of dendritic spines in learning.
Chapter five is devoted entirely to the "old" brain, the one that is not the neocortex. The sensory information pathways, the role of the thalamus, what the hippocampus is, the role of dopamine and serotonin for fear and pleasure, basically what we’ve been discussing Interesting, but very succinct. As a conclusion to the chapter – for the tasks described above, we don’t need to learn the old brain at all. Well, he knows best.
Chapter six is a short chapter on human ability, creativity, and love. It’s really about nothing. If you want to learn interesting things, read AlexeyR
And so, by chapter seven, we’ve finally arrived at the stated in the title : the digital neocortex from biology. A lot about the history of computer modeling of the brain, where neural networks came from, the role of classifiers and hidden Markov models, a bit about genetic algorithms and how they should be applied, a bit about Lisp and Jeff Hawkins. A lot about Watson and Jeopardy!, but not how it works (a little bit about that at all), but how cool and important it is! A whole section "Brain Building Strategy" summarizing the methods and how to apply them and what else would be nice to add. Very descriptive chapter, almost no specifics.
Chapter 8 – The Brain as a Computer. A chapter on Thuring, von Neumann, and Bebbidge. That’s what I learned in computer science at school. How about you?
And then the philosophy begins: Chapter 9 – Mental Experiments with the Brain. Here we learn what consciousness is through the examples of zombies and qualia about Penrose and his quantum theory of consciousness (Ray doesn’t approve of him), faith, the difference in how Easterners and Westerners perceive the world, free will, the determinacy/predictability of life, cloning consciousness. Very interesting topics for civilized conversation among friends, but the stated topic was about the creation of the brain, not how this brain fits into our reality.
Chapter 10 in this book was added purely because Ray likes to make predictions :). It’s called the law of exponential growth of technology (loosely translated). And he discusses his old predictions from the book The singularity is near.
Well and chapter 11 finally is objections, i.e., what other people are saying about his approach. And most of the chapter is devoted to Paul Allen. Yes, that one. They’ve had a long-standing feud. Paul turned out to be wrong on all sides.
Here’s a little book like this. Very artistic. Practical value, especially for those who, for example, took an online AI course at Stanford, tends to zero, but for general development as a bedtime book is fine.

You may also like