le 16 novembre 2012
*A full executive summary of this book will be available at newbooksinbrief dot com on or before Monday, November 26.
When IBM's Deep Blue defeated humanity's greatest chess player Garry Kasparov in 1997 it marked a major turning point in the progress of artificial intelligence (AI). A still more impressive turning point in AI was achieved in 2011 when another creation of IBM named Watson defeated Jeopardy! phenoms Ken Jennings and Brad Rutter at their own game. As time marches on and technology advances we can easily envision still more impressive feats coming out of AI. And yet when it comes to the prospect of a computer ever actually matching human intelligence in all of its complexity and intricacy, we may find ourselves skeptical that this could ever be fully achieved. There seems to be a fundamental difference between the way a human mind works and the way even the most sophisticated machine works--a qualitative difference that could never be breached. Famous inventor and futurist Ray Kurzweil begs to differ.
To begin with--despite the richness and complexity of human thought--Kurzweil argues that the underlying principles and neuro-networks that are responsible for higher-order thinking are actually relatively simple, and in fact fully replicable. Indeed, for Kurzweil, our most sophisticated AI machines are already beginning to employ the sample principles and are mimicking the same neuro-structures that are present in the human brain.
Beginning with the brain, Kurzweil argues that recent advances in neuroscience indicate that the neocortex (whence our higher-level thinking comes) operates according to a sophisticated (though relatively straightforward) pattern recognition scheme. This pattern recognition scheme is hierarchical in nature, such that lower-level patterns representing discrete bits of input (coming in from the surrounding environment) combine to trigger higher-level patterns that represent more general categories that are more abstract in nature. The hierarchical structure is innate, but the specific categories and meta-categories are filled in by way of learning. Also, the direction of information travel is not only from the bottom up, but also from the top down, such that the activation of higher-order patterns can trigger lower-order ones, and there is feedback between the varying levels. (The theory that sees the brain operating in this way is referred to as the Pattern Recognition Theory of the Mind or PRTM).
As Kurzweil points out, this pattern recognition scheme is actually remarkably similar to the technology that our most sophisticated AI machines are already using. Indeed, not only are these machines designed to process information in a hierarchical way (just as our brain is), but machines such as Watson (and even Siri, the voice recognition software available on the iPhone), are structured in such a way that they are capable of learning from the environment. For example, Watson was able to modify its software based on the information it gathered from reading the entire Wikipedia file. (The technology that these machines are using is known as the hierarchical hidden Markov model or HHMM, and Kurzweil was himself a part of developing this technology in the 1980's and 1990's.)
Given that our AI machines are now running according to the same principles as our brains, and given the exponential rate at which all information-based technologies advance, Kurzweil predicts a time when computers will in fact be capable of matching human thought--right down to having such features as consciousness, identity and free will (Kurzweil's specific prediction here is that this will occur by the year 2029).
What's more, because computer technology does not have some of the limitations inherent in biological systems, Kurzweil predicts a time when computers will even vastly outstrip human capabilities. Of course, since we use our tools as a natural extension of ourselves (figuratively, but sometimes also literally), this will also be a time when our own capabilities will vastly outstrip our capabilities of today. Ultimately, Kurzweil thinks, we will simply use the markedly superior computer technology to replace our outdated neurochemistry (as we now replace a limb with a prosthetic), and thus fully merge with our machines (a state that Kurzweil refers to as the singularity). This is the argument that Kurzweil makes in his new book 'How to Create a Mind: The Secret of Human Thought Revealed'.
Kurzweil lays out his arguments very clearly, and he does have a knack for explaining some very difficult concepts in a very simple way. My only objection to the book is that there is a fair bit of repetition, and some of the philosophical arguments (on such things as consciousness, identity and free will) drag on longer than need be. All in all there is much of interest to be learned both about artificial intelligence and neuroscience. A full executive summary of this book will be available at newbooksinbrief dot com, on or before Monday, November 26; a podcast discussion of the book will be available shortly
le 26 janvier 2013
We must understand this title that pretends to tell you how you can create a mind has to be taken literally. Ray Kurzweil believes in his Artificial Intelligence engineer’s enthusiasm that he can create a mind, that he may qualify as god himself, a secular god as a matter of fact.
“Evolution can then be viewed as a spiritual process in that it creates spiritual beings, that is, entities that are conscious. Evolution also moves toward greater complexity, greater knowledge, greater intelligence, greater beauty, greater creativity, and the ability to express more transcendent emotions, such as love. These are all descriptions that people have used for the concept of God, albeit God is described as having no limitations in these regards.” (p. 223)
And do not consider all that is pure rhetoric or pulpit preaching. He believes evolution is the real God when he says: “Our neocortex is virgin territory when our brain is created . . . the biological process of actually growing a brain.” (p. 62) We can wonder about this evolution or biological process if it is a creator or a grower, God or a simple farmer. But we have to wonder what Kurzweil means by “brain” and “mind.” Page 23 over 26 lines he uses the following string of words: “mind . . . brain . . . mind . . . theories . . . ideas . . . thought . . . thinking . . . theories . . . thought . . . brain . . . thinking . . . “ We can assert that these words are not really discriminated. This lack of clear definitions of these terms is of course an enormous shortcoming that is just as nearly irritating as the levity with which he deals with Einstein: “Einstein articulated my goals in this book well when he said that ‘any intelligent fool can make things bigger and more complex . . . but it takes . . . a lot of courage to move in the opposite direction.’” (p. 11) It is obvious Einstein did not articulate his goals since he has not been alive for a while now. That use of the passive by Kurzweil to draw to himself what the quoted person said is even more astounding with at least two and quite often more than three quotations, at times long ones, at the head of all chapters and even subchapters. Kurzweil seems to forget that quoting does not prove anything. But this quoting and bringing together opposed ideas is the basic unitarian objective of the author:
“The truth can be discovered only by finding an explanation that overrides – transcends – seeming differences, especially for fundamental questions of meaning and purpose. That is how I resolve the Western-Eastern divide on consciousness and the physical world. In my view both perspective have to be true. On the one hand it is foolish to deny the physical world . . . On the other hand, the Eastern perspective – that consciousness is fundamental and represents the only reality that is truly important – is also difficult to deny.” (p. 222)
On one hand blunt and brutal materialism since Kurzweil does not seem to consider the material existence of the mind, except when reduced to the brain, or of ideas, thoughts, ideologies, etc. On the other hand a principle that is derived from a false reference to Buddhism.
“In the Eastern view, consciousness is the fundamental reality, the physical world only comes into existence through the thoughts of conscious beings . . . I call this the Buddhist school of quantum mechanics, because in it particles essentially don’t exist until they are observed by a conscious person.” (p. 218-219)
Kurzweil does not know what he is speaking of. Buddhism is basically expressed in the Dhammapada and the Abhidhamma. For Buddha the whole material world exists outside our consciousness and we are part of it because we have a body. This whole world can only be captured by our six senses, the five basic senses plus the mind as a meta-sense that processes the sensations captured by the five other senses plus the abstract concepts conveyed by language and organized in abstract reasoning or description. The word “consciousness” that Kurzweil uses does not correspond at all to the words used for the “mind” that sixth sense or meta-sense. In fact there are two words in Pali for the mind, “mana” that refers to the meta-sense itself and “citta” which refers to the various mental states of an individual experiencing some type of feeling, emotion, mental excitation, etc. Kurzweil uses the word “determined” a lot about the material world. There is a Buddhist concept behind. The whole physical world, including us as physical beings is determined, follows the physical laws governing the cosmos. By using the mind any individual can get into meditation, which will lead him onto the eightfold path of illumination that is to say the possibility to get detached from the determined world and hence to merge with cosmic energy once death has come, thus getting out of the triple characteristic of the determined world: everything is changing all the time; everything is carried by a cycle that goes from birth to life and decay then to death and then to rebirth. Nibbana (known in Sanskrit as Nirvana) is that mentally produced escape from this cycle into cosmic energy; everything has no essence, soul or permanence of any type.
This is important because this should lead us to refusing the basic objective Kurzweil gives to humanity: to use intelligent machines to “coloniz[e]” (p. 281) the universe. In previous books he was rejoicing in the idea that the speed of light could be stepped over, hence speeding the “colonizing [of] the universe” (p. 281) though in this book he is more realistic since the good news about having transported molecules at a speed higher than the speed of light has been disproved in this very 2012 year. But the objective remains: to colonize the universe. Some people never learn. The colonization of the planet by the Europeans has not exactly been the best thing in the world producing slavery, the eradication of American Indians, Aztecs, Mayas, Incas, etc, colonialism and throwing three continents, if not four into, underdevelopment and exploitation. It is high time Kurzweil questions his basic fundamental motivation. The conquest of the universe is not on the agenda. So far we are dealing with the discovery of the universe. We might never conquer it, especially if intelligent beings exist here and there. The use of the cavalry seems to be slightly passé.
This said, and it is fundamental we can move to the main subject of the book: the mind, though in fact he never speaks of it reducing it to the brain. So let’s start with the brain.
After a rather long career and many books published on his “Singularity” that was and still is heftily criticized by many people in the field, including people who are specialists, theoreticians and entrepreneurs in computing science and technology like Kurzweil himself, he wrote this book to get back in phase with others. Criticism was generally rejected high-handedly before. This time he makes an effort to integrate the research of others in the first half of his book, hence to describe the functioning of the brain the way it is known by scientists, though in the second half of the book he goes back his messianic, apocalyptic, prophetic, oracular prediction of the merging of biological intelligence, hence man, into non-biological intelligence, hence machines and we jump onto the track to Terminator 25 all over again and dreams of a time when “computers will have . . . surpassed unenhanced human intelligence.” This phrase gives us in a nutshell, not a walnut but a hazelnut, his basic thinking. Note he of course neglects the fact that human intelligence develops along with all the intelligent machines and theories man has invented. If these intelligent machines are used properly, that is to say at the top of their capabilities, then the intelligence of the users will tremendously develop. Will we have a new mutation in biological evolution? Some human beings are able to develop some tremendous capabilities as for memory, the assimilation of hierarchical systems like foreign languages, etc. These are supposed to be autistic, but do we know anything serious about autistic people apart from believing they are different and have to be put away?
Let’s speak of the brain now. I will not be over technical about it. He borrows from various other researchers (Jeff Hawkins, Dileep George and Jaron Lanier mostly) the general architecture of the brain and adds a couple of things.
The neocortex is the part of the brain that controls our most advanced human intelligent activities. It has six layers and it is structured in vertical columns across these six layers; Each column hence has six layers too. These columns are connected in many ways first of all to the columns around each one of them on a proximity basis, but some spindle neurons can connect many columns in all parts of the brain, 60% of these spindle neurons in the right hemisphere and 40 percent in the left hemisphere. They appeared with hominids, our ancestors after branching out of apes some 10 or 15 million years ago. But we must know that they already existed in apes since Gorillas have about 20% of our number, Bonobos have 2.5% and chimpanzees about 2%. Other mammals do not have any at all. Kurzweil does not speak of mirror neurons and he should have since they are also only vastly present in Homo Sapiens, though they must have been present in hominids and are present in some apes, and these are essential for learning and empathy since they enable someone to imitate the actions of someone else and to empathetically feel the same emotions as other people around them. He also mentions though lightly the fact that a fetus has a brain as soon as one month of age and this fetus will hear (he does not mention this one) and see around the 20th or 24th weeks of pregnancy. He forgets to say that the brain grows after birth. But he does mention that everything happening while the brain is growing has important consequences on the growth of this brain. But he makes his basic mistake here at the very basis of his approach.
First he considers that “learning and recognition take place simultaneously.” (p. 63) He just forget in the womb the fetus cannot learn because all he hears or feels has no referential dimension; These sensations he feels and the sound clusters of any type he hears are registered, that was proved, but with no reference, hence no real meaning, though they can have a comforting or disturbing effect on the fetus along with the mother’s mood. After birth it is obvious then the baby has the possibility to attach a referent hence a meaning to what he sees and identifies. At this point it is impossible to say that learning and recognizing happens simultaneously for the same things. You have to learn about something before recognizing it. Even if is only a comforting sensation you have to experience it first, to more or less identify it second before being able to recognize it. Recognition is necessarily second at least because to identify you have to experience several times and that’s what he probably means. The first time you just experience, the second time then you recognize and by recognizing you identify even if it is superficially. But there must be a first moment of pure experience. But this is nothing in itself. The main shortcoming at this moment is the absence of any consideration about language. For Kurzweil language, spoken first and written second are the only two inventions of humanity (he says so twice p. 27 and 159) bringing together in one movement two human inventions that have at least 300,000 years between them and it neglects the phylogeny of that linguistic ability. Once again without entering details, language which was oral only for at least 300,000 years out of 305,000 years is an invention of humanity, ,hence of the brain and since language is not something you can touch it is part of the mind. Written language will only come very late in human history. There are still some human groups on the earth that do not write at all.
To invent human articulated language the neocortex has to have a hierarchical organization, which is the case in each column and in the neocortex all together and within the brain between the old brain and the neocortex. That hierarchical architecture of the brain makes the brain only able to function along that line. The hierarchical architecture of the brain produces hierarchical thinking, hierarchical language, hierarchical society, etc. All human activities contain a hierarchical dimension that is the reflection of the architecture of the brain. And here with language you hold an essential line of thought. Every single advancement in phylogeny, in lexicon, in syntax is produced by the mind and each advancement is inscribed in the mind and determines the next advancement. We could show how complex but also how direct and simple this transitive productive process is. What’s more the experience of a human being in front of any entity is hierarchical. He must first discriminate it. Then he has to identify it and name it with a new name if it is a new entity or an old name if he recognizes it as already known and named. Then it has to be classified and that leads to another abstract operation that is known has conceptualization. There is no concept if there is no conceptualization; Kurzweil uses the word “concept” several times, though he does not list it in his index, but he does not use the word “conceptualization” which means for him concepts are generated by magic.
It is obvious then that written language amplifies the intellectual conceptualization of people since they do not have to simply remember plain facts that are recorded in books. They can step further into more abstract thinking. Imagine what it is when you have the Internet at the tip of your fingers. There are thus systematic hierarchies that he neglects. From root to theme and then frond at the level of the semantic units of the language often called words. From syncretic concatenation, to clause structure, to multi-clause structure by concatenation and then embedding, as for syntax often called grammar. From simple calls, to orders, to descriptive discourse, to explicative discourse, to any other discourse with an ever higher level of abstraction, social meaning, content or intention, and that has to use various media to be uttered or produced.
But there is more if you cross brain and language.
Each column, and that is Kurzweil’s approach, is composed of many modules, each one having about 100 neurons. These modules are connected inside the column in complex intertwined networks. Hence we then have a first hierarchy: neurones and their relations within a module, then modules and their relations within a column and at each level relations between the elements and the direct outside: neurons from one module to neurons from another module in the same column, modules from one column to modules from the same column and to modules from other columns. And yet we miss the spindle neurons that can connect any column to any other column and any module to any other module. These spindle neurons seem to be totally opportunistic and develop according to the needs of this or that moment for this or that individual. We thus get to what Dileep George calls “recursive cortical networks” (quoted p. 152) and I insist on the fact that these networks are growing from nearly conception to death, or at least to an advanced age, as long as we can learn new tricks, that they are flexible and versatile in many ways, which explains why we can learn new things all the time: there is plenty of room in the brain and any learning does not depend on a type of available neurons, they are all basically the same. This enables man to use many ways of thinking and one at least is unpredictable and hence inimitable.
The simplest way is to put together two entities and their proximity implies they are connected. It’s what is called syncrertic thinking and it corresponds to what Kurzweil calls “leakage” in the brain, one neuron being in a certain state due to some influx of information coming to it may have a direct influence on its neighbors as if it leaked its information over his surrounding neurons.
Then we can build a deductive argumentation. One event is the cause of another which is the effect of the first one. We can thus build deductive chains. That’s the standard reasoning in sciences like mathematics. We can also inverse the reasoning and get into an inductive chain of reasoning. From what I know I induce that this should be true. It is a hypothesis. This is also important in sciences, but also in everyday life like: it was raining yesterday hence my father must have stayed home.
But there is another way of thinking. I call it subduction. The simplest form of subduction is a metaphor; I treat one entity as if it were another and that may reveal an aspect of the first entity I had not seen at first. A metaphor or a subduction does not prove anything. It has to be demonstrated afterwards, but that’s how the most creative activities of man develop. We have a deep feeling, a strong emotion, a profound conviction, post traumatic stress, and we draw from this the idea that the working truth should be this or that. It is an induction in a way but a lot vaster and deeper. This sudden truth is the Eureka of Archimedes. A sudden illumination. Note such epiphanies can happen at any time and anywhere and in any field of activity. This subduction corresponds perfectly to the recursive cortical networks Dileep George is speaking of. Note language is not indispensible. A composer can just experience such epiphanies in his composing and he would be unable to explain in words what it means. That’s generally why I would consider the artistic creator is the last person who can explain his own creation.
The question that I will only evoke here is where do these elements of the mind of a person register in the brain. Kurzweil does not even ask the question. For him whole lists of patterns as he calls them are available in the various modules. The question is to know where all these elements, patterns or not, are registered. My idea is that we are working at the level of the molecules with particularly the proteins in the microtubules of the neurons since it is proved some of these proteins can vary including in structure when impacted by some influx of information. Same thing about the transportation of the sensorial information from the sensorial organs to the brain: how is it done? A vast discussion is needed here.
The last point I would like to make here is about Artificial Intelligence. Kurzweil’s objective is to copy a real brain, or maybe several (though mixing two brains might produce strange effects since there cannot be two brains that are identical due to their psychogenetic history), and then compress the information by cutting out all redundancy and the brain is very redundant. Kurzweil says that should have no effect. I would doubt it since each instance of one piece of information was registered in one specific situation with particular emotional or sensorial elements around it and these variations from one recording to the next of the same item will be lost by compressing. Then he will simulate that compressed version of a brain in an intelligent machine. His machine will only be able to simulate the compressed version of one particular brain and hence will in no way represent the human brain at the level of its abstract totality. But Kurzweil knows it is in many ways bound to be too short:
“Almost certainly we would not find a precise match; the neuronal structure would invariably differ in many details compared with the models in the computer. However, I would maintain that there must be an essential mathematical equivalence to a high degree of precision between the actual biology and our attempt to emulate it; otherwise these systems would not work as well as they do.” (p. 153)
What is lost in such a simulation is what makes a brain different from all others, the circumstantial elements attached to each item of knowledge, but it is these elements that may be particularly pervasive in a subductive inspirational way of thinking. A plane after all flies pretty well but it is quite different from a bird, isn’t it, though it performs the task of flying quite well.
Dr Jacques COULARDEAU