“How to Create a Mind”
2012, Viking Press, 352, $10-$16
Merriam-Webster dictionary defines the word “transcendent” as “being beyond the limits of all possible experience and knowledge”, and naturally, the word “transcendence” as being the act thereof. It is notable that the definition of the word itself directly states that the act is “beyond…possible” and therefore not within the ability of any human or other being to achieve. The movie Transcendence, currently playing in theaters, boasts this unusual word as its title in order to accurately depict the equally-unusual and thought-provoking plot the movie promises in its trailer. Set in modern times, writer Jack Paglen foretells of an alternate present in which Dr. Will Caster has discovered a way to create a super-intelligent machine which can see, hear, feel, touch, and even interact with its environment, while at the same time having the full emotional capacity of the human mind. By doing this, Dr. Caster makes it possible for a person’s consciousness and mind to be uploaded onto his machine, and as one might be able to predict, that is exactly how the movie progresses.
The concept of creating artificial intelligence (A.I) in today’s day in age is not out of the realm of plausibility; in fact, many would argue that mankind has invented a number of different computers and machines which would, by most standards, be considered intelligent. But it is the concept of a machine which has artificial consciousness that still seems to be limited to the classic works of Isaac Asimov and Alan Turing (arguably the two most influential people in the development of A.I.; in both actual research and in its fictional presence). It’s been 75 years since Asimov began writing his influential Robot Series short stories, and 60 years since Alan Turing published his paper on the Turing Test and yet, the field of artificial intelligence is still largely underdeveloped.
In November of 2012, Ray Kurzweil, a renowned inventor, scientist, and self-described futurist, published the book How to Create a Mind which set out to, as his subtitle suggests, “reveal the secrets of human thought”. To say that this book had a rocky reception among Kurzweil’s peers may be a bit of an understatement. Gary Marcus, a research psychologist who, much like Kurzweil, has spent a great deal of time studying not only the human brain, but the mind itself, criticizes the book in some of its most basic levels (Marcus, 2012). Kurzweil’s “Pattern Recognition Theory of the Mind” which he cleverly abbreviates to simply “PRTM” seems to be one of the key points Kurzweil attempts to sell his audience throughout the book; however it also seems to be the biggest source of criticism from others in his field. Marcus states in his analysis that Kurzweil’s PRTM is actually not a new theory which Kurzweil should be taking credit for, because of its incredible similarity to the neocognitron model. In Marcus’s own words, “Anyone who knows the history of A.I. will recognize that the basic theory (and even the diagrams that are used to illustrate it) is very much in the spirit of a textbook model of vision that was introduced in 1980, known as neocognitron”. Moreover, Marcus feels that the PRTM also resembles the Hierarchical Temporary Memory (HTM) model developed by Jeff Hawkins and that any difference between the two would not constitute the need for yet another acronym. To add insult to injury, Marcus keys in on one more fundamental issue with Kurzweil’s logic; the use of “thought experiments”. Throughout the book, Kurzweil uses the term “thought experiment” to classify simple brain-game-like tests of basic logic and assumption which he believes to be sound evidence to back his claims. Although Kurzweil occasionally provides a smidge of outside evidence to back up his points, much of the book is reliant on these so called “thought experiments”. The issue with this way of thinking is that it leaves far too much uncertainty and guess work in the argument. In Marcus’s own words; “Not a single cognitive psychologist or study is referred to, and he scarcely engages the phenomena that make the human mind so distinctive” (Marcus, 2012).
Marcus is not alone in his criticisms of Kurzweil’s thesis and primary points of the book. Many other critics such as Simon Garfinkel, and Colin McGinn (a computer scientist and philosophy professor respectively) also express their concerns with the logic Kurzweil uses in his book, and all mention the apparent lack of evidence he provides for his rather grandiose claims. I personally can attest to the fact that even to the common reader, the lack of evidence for Kurzweil’s claims is noticeable and I found myself constantly questioning what rationale he had for determining what the ‘correct’ way of thinking was. This style of writing is unfortunately one which I sensed throughout the entire book, as did Garfinkel, Marcus, and McGinn. However, I don’t feel that How to Create a Mind is a completely useless book; far from it.
The dictionary definition of consciousness, “perceiving, apprehending, or noticing with a degree of controlled thought or observation”, gives a relatively broad concept of what consciousness can be described as. One of the issues with defining consciousness is deciding which existing beings we already consider to be conscious. Most humans would probably agree that they themselves are conscious due to the fact that they can think their own independent thoughts and operate with a degree of free will. Likewise, most humans assume that the people around them are conscious under the same reasoning. But is this leaving too much to assumption? How do we know with certainty that the people around us are not simply programmed to act the way that they do? To a degree, the aforementioned idea must have some truth to it. After all, our consciousness is merely a product of our functioning brain and way of thinking, and therefore there must be something in our brains that is so wildly complex that the product of our brain’s wiring establishes what we perceive as consciousness. If this is the case, then it would seem to be entirely possible, on a theoretical level at the very least, to create a machine which is conscious. How this would be done is really what Kurzweil’s book was supposed to be about – though the critical reviews of the book may suggest that it failed at this.
Kurzweil’s approach to this ever daunting task is to use the human brain as a base model for creating an artificial one. The problem with this plan is that Ray Kurzweil is only a computer scientist. He is not a psychologist, a neurologist, biologist, nor a philosopher; he knows computers, and old ones at that. It is apparent to me that the project of creating an artificial human mind is not one which can be solved in the field of computer programming and artificial intelligence alone, but rather that a number of different fields (such as the ones listed above) may be needed in order to fully understand the breadth of this project and the question it asks about our own humanity. It would therefore only make sense to consult experts in each of these fields at each stage of the scientific process; starting with the basic question “What defines consciousness?”
Christopf Koch and Giulio Tononi (both well respected Neuroscientists) may have started the ball rolling as they attempt to answer this very question in their 2008 essay Can Machines Be Conscious (Koch & Tononi, 2008). Koch and Tononi are able to ultimately conclude that consciousness is not defined by the ability for one to engage with its environment; one does not need memory, nor be attentive, nor have self-reflection; one does not need a language, nor emotion, and that a machine would therefore not need any of these things in order to be considered conscious. Instead, they concluded that the distinguishing factor between a conscious being and an unconscious being is the amount of data it perceives and what that data means to the subject. To refine their definition even further, they specified that the data it receives cannot be a series of subdivided information and that all of the information the being receives must be received collectively as one image or as one memory. In their own words:
The camera can indeed be in any one of an absurdly large number of different states. However, the 1-megapixel sensor chip isn’t a single integrated system but rather a collection of one million individual, completely independent photodiodes, each with a repertoire of two states. And a million photodiodes are collectively no smarter than one photodiode. By contrast, the repertoire of states available to you cannot be subdivided. You know this from experience: when you consciously see a certain image, you experience that image as an integrated whole. No matter how hard you try, you cannot divvy it up into smaller thumbprint images, and you cannot experience its colors independently of the shapes, or the left half of your field of view independently of the right half. (Koch & Tononi, 2008)
A major difference seen in this paper is the much more prevalent use of citing research and studies which prove the point the two are trying to make in their paper. Additionally, they explain in detail which parts of the experiment they are actually discussing and how it fits in with their point. They go on to talk about the logistics involved in creating a conscious mind, citing more research and experiments, many of which are still ongoing; ultimately concluding that there is a lot more work to be done. Although their paper does not give a specific date as to when we could expect to see an artificial consciousness, they give mention that it would take more than a few decades to complete the neurologic research alone. As for Kurzweil, he feels that such conscious beings will find their way into our lives around the year 2029 “and become routine in the 2030s” (p.210). In saying this, Kurzweil brings up an entirely different point, which is whether or not people want to create an artificial consciousness, and what the uses for one might be. Koch and Tononi don’t explore this idea to any great length, with the exception of their conclusion, in which they optimistically mention that it would give us more insight into how our own consciousness works. I feel that the ideas and overall project that Kurzweil attempts to tackle in his book is of major importance to the fields of artificial intelligence, neuroscience, psychology and numerous others. However, it seems to be equally true that one scientist, from any field or academic background, will not be able to tackle this project on his or her own. It would take a panel of such people, a large amount of grant money, and a very long time before we would be able to say with some shred of certainty that a conclusion as to what the definition of consciousness is can be proven scientifically. Assuming not only that it can be proven scientifically, but also that such a panel was able to reach a conclusion, then it would take many more years and more grant money to tackle the more complicated and diverse question of how an artificial consciousness would be assembled.
Koch, C., & Tononi, G. (2008) Can Machines Be Conscious. IEEE Spectrum. Retrieved from http://spectrum.ieee.org/biomedical/imaging/can-machines-be-conscious
Marcus, G. (2012) Ray Kurzweil’s Dubious New Theory of Mind [Review of book How to Create a Mind]. The New Yorker. Retrieved from http://www.newyorker.com/online/blogs/books/2012/11/ray-kurzweils-dubious-new-theory-of-mind.html