Pulse en una miniatura para ir a Google Books.
Cargando... The Myth of Artificial Intelligence: Why Computers Can’t Think the Way We Dopor Erik J. Larson
University Presses (154) Cargando...
Inscríbete en LibraryThing para averiguar si este libro te gustará. Actualmente no hay Conversaciones sobre este libro. sin reseñas | añadir una reseña
Futurists are certain that humanlike AI is on the horizon, but in fact engineers have no idea how to program human reasoning. AI reasons from statistical correlations across data sets, while common sense is based heavily on conjecture. Erik Larson argues that hyping existing methods will only hold us back from developing truly humanlike AI. No se han encontrado descripciones de biblioteca. |
Debates activosNingunoCubiertas populares
Google Books — Cargando... GénerosSistema Decimal Melvil (DDC)006.3Information Computer Science; Knowledge and Systems Special Topics Artificial IntelligenceClasificación de la Biblioteca del CongresoValoraciónPromedio:
¿Eres tú?Conviértete en un Autor de LibraryThing. |
So when all the hype started couple of years ago I was taken aback. For any question I asked I got no answer - be it verbal or in texts in magazines. Everything came to - it works on its own, it is just required to add sufficient data. OK, this sounds like expert system but what about AI, how does it reason? Same answer. And it was given to me in a way like I was most stupid man ever for not getting it. And I started feeling like that after every discussion with my colleagues - AI is here, will take whole bunch of tasks because you can program it to do almost everything. OK, all good, great, but bloody how? Just add more data.
And this put me in place where I could not make sense out of anything. I was sure I was missing something and tried finding additional literature - unfortunately these were such a mish-mash of wishful thinking that it only left me with more questions.
And then I found this book.
In a very concise way author gives the overview of current AI research and its rather sorry state.
Author is very to the point and he writes as if he was asked so many times about the AI [by people like me :)] that he decided to write a book to provide reference to everyone interested in the field.
Book is somewhere between popular and mid level science book. Few chapters that deal with logic and rules of logical reasoning might be uninteresting to people that are not in field of computer science or applied mathematics but rest of the book is accessible to everyone.
And what a damning picture this book paints. Presented with the possibility that there are no more major breakthroughs ahead (rather theoretical Nobel prizes notwithstanding I think there is still lot more practical research to be done) scientists took an unscientific approach - instead of research they changed their course and joined forces with the corporations. Corporations, being in their nature, decided to push their own products as they are, because any further research would cost, and marketing hype came in force. Result? Terrible. It only confirmed that for person with hammer every problem looks like a nail.
Due to the hype that came form all known authorities in the field (corporations and individual scientists associated with them) states started funding research that was using AI (in truth only classification engines) and neglecting others, that went more traditional way. As a result AI research was a major hit - nobody wanted to "waste time" when AI will definitely bring results (and if anyone asked how, answer was definitely "Add more data").
As author clearly states - entire AI research to date is based on constant futile attempts of simplifying human brain to the level of chain of data processors. Futile because this same attempt failed over and over again (including now, even with the enormous computer power and data collections). Reason? Very simple, how can one build intelligence when we do not know anything about our own intelligence process (when I read this I was stupefied, even after all these years we still do not have definitive answer on how our mind works, mind-blowing).
It is like entire science community decided to decipher how does automobile work by just looking at the outside parts and not being aware of the main part - engine. And expected good results. Blimey.
As a result complete AI research wasted good part of last decade. Significant improvements were made for classification systems and expert systems (as author says very very narrow AIs) but everything else was stopped in its tracks.
Unfortunately hype caused quite a social upheaval. I agree with the author, it looks like anti-human revolution took place, humans were discarded as used tools (it is incredible how this morbid cult of human irrelevance and expendability took root world wide in last decade) all in expectations of rise of our machine masters. Benevolent or not, it seems it does not matter to any of the technocratic leaders - they are so eager to give birth to something - without even knowing what exactly.
In a short time use of classification engines helped to create a divide between people by pushing news people like to read. It is not their error, mind you, they do what they are programmed to do but inadvertently this back-fired because of society totally surrendered itself to these computer idiot-savants for everyday news and information - from food to politics.
This is very timely book, and I hope that author's message to bring back sense in AI scientific research is accepted by the community. AI in any form can do wonders for humans but it must not be goal in itself. It is a device that can bring enlightenment and propel the humanity forward, but that can only be achieved without trickery [in scientific approach] and by following age old scientific approach (getting to theories and proving or disapproving them] that proved itself many times over.
Will it take time? Definitely. But this will help us to perform detailed and valuable research and, which might be more important, we will become mature enough to cope with the end-results.
Excellent book, highly recommended. ( )