Kurzweil's PRTM Next Step in AI
Persian Version متن فارسی
Ray Kurzweil's new book How to Create a Mind about Artificial
Intelligence (AI) has just been published (1). The book examines topics that
have long been the subject of discussion in the AI community (2).
Nowadays one does not need to discuss writings such as Turing's monumental 1936 paper to draw attention to AI (3). Anybody making a query on Google wants the search engine to respond like a human, i.e. to understand what the inquirer is asking and to use the vast information it has at its disposal to respond appropriately. People are not afraid of such engines; rather, they wish it would work like a superhuman. In other words, after over half a century, AI is increasingly implemented in tools like Google search engine, albeit with its current limitations, and that luddites are disappearing at least in this area.
Kurzweil in his new book goes back to the basics which have been mostly missing in the literature of the field for two decades. When the word AI was coined by John McCarthy of Stanford, two distinct views of AI were easily discernible. The first was that of McCarthy himself. McCarthy, the inventor of the LISP language that was widely used by AI practitioners in the early days, viewed human reasoning as a non-monotonic logical system. Simply put, this means that if we say a dog is an animal with a tail and if later we see a dog that does not have a tail, then in this logical model, we consider the new case as an exception.
In contrast, Marvin Minsky of MIT, who is considered the founder of AI, had a different view of intelligence. He proposed a theory of frames, which was expounded in his monumental book Society of Mind, published in 1988. According to Minsky's view, when we say a dog is an animal with a tail, we create a frame in our brain corresponding to this understanding and once we run into a case where a dog does not have a tail, we just modify the frame.
In his new book, Kurzweil proposes a new theory to describe human reasoning called Pattern Recognition Theory of Mind (PRTM). Kurzweil shows the crux of his theory by writing that "each of our routine procedures is remembered as an elaborate hierarchy of nested activities" (4). He gives the excellent example of how easy it is for us to remember the alphabet in its sequence but how hard it is for us to recite it in reverse (5). His PRTM model is not speculative but rather is based on his extensive work on reverse engineering the brain. Thus it surely can be used by AI practitioners to better simulate human thinking.
I believe Kurzweil's model is based on how the brain of most abstract thinkers works, referred to by NLP theorists as cerebral people. Neuro-linguistic programming (NLP), a field founded by Richard Bandler and John Grinder, is not viewed as a science in many academic circles, and some personal issues that came up early in that field made it especially controversial. Ignoring the psychology-related claims made by NLP, I believe it offers a structure where we can see that not all human brains work the same. For instance, when Kurzweil describes how he remembers seeing a woman pushing a baby carriage during his morning walk, he writes, "I cannot recall what either of them was wearing or the color of their hair. (My wife will tell you that this is typical.) Although I am unable to describe anything specific about their appearance, I do have some ineffable sense of what the mom looked like" (6).
The reality is that visual people do not need to pay attention to the details of someone's appearance, but when asked to remember those details, they may easily retrieve them. By contrast, non-visual people must pay attention to those details to remember them later. Maybe the former group has what is sometimes referred to as photographic memory when at its best. At any rate, most people referred to by NLP theorists as 'visual' people think this way, but those who are referred to as 'auditory' or 'kinesthetic' people happen not to think this way. Cerebral people may belong to any of these three groups. Now would these factors make a difference to Kurzweil's PRTM when he tries to reverse engineer the brain? It does not seem to be the case, because he is focusing on pattern recognition regardless of whether the patterns are perceived visually, auditorily or kinesthetically. Thus the methods described in his book should not change even when considering this difference. Nonetheless this factor may change how we explain some of the research work that Kurzweil has presented in his book. This is why I think that further research should be conducted on the mechanisms by which visual people are able to remember those aforementioned details without having paid attention to them and which are missing in non-visual people.
I found Kurzweil's How to Create a Mind to be an excellent exposition of AI as it is today with all of its potentials, and after decades of not having a treatise of fundamentals in the field, it was refreshing to see such a masterpiece published by Kurzweil, who is the most pioneering thinker of the field in our times.
Sam Ghandchi, Editor/Publisher
November 19, 2012
1. Raymond Kurzweil, How to Create a Mind, Viking, 2012
2. Understanding Self-consciousness: Differentiating Humans from other Sentient Beings
3. Turing and AI
4. Raymond Kurzweil, How to Create a Mind, Viking, 2012, Page 33
5. ibid, Page 27
6. ibid, Page 29