Comment Re:AI is a fraud (Score 2, Insightful) 390
Did you work in the lab as a guinea pig, by any chance? Because you seem to have very little understanding of what AI is about.
The Turing test is not relevant for one simple reason - it does nothing to connect the lines of research that are currently being actively (and - gasp!- successfully, on occasion) pursued in the field of AI. Yes, fifty years ago AI researchers thought that computers would be able to mimic the behavior of human beings: but the field was just emerging at the time, and people didn't have an understanding of the tremendous complexities involved. (Consider, for comparison, that at the time calculus emerged, people thought that it would be the ultimate tool for explaining all mysteries Life, Universe, and Everything.) Since that time however, these complexities became apparent - many problems in AI have been shown to be Turing-undecidable (no algorithmic solution is possible), and virtually every open problem in the field is at least NP-hard. Furthermore, you cannot really develop human-like intelligence without understanding how the human mind works, and that understanding is sadly lacking. It took hundreds of millions of years for evolution to produce intelligent beings: and this is intelligence is manifested through a massively parallel, largely mysterious mechanism (the brain). Is it even remotely reasonable to expect that researchers would be able to emulate this poorly understood mechanism using vastly different underlying hardware in only 50 years?
The point - one that you are so throughly miss - is that the goal of AI is not to build cute robot friends for the human race, but to devise complex systems to solve specific problems that currently require human intelligence. As such, the field of AI has long ago been split into intersecting sub-domains of machine learning, knowledge represetnation, planning, natural language understanding, e.t.c. Each of those has had some degree of success, and produced plenty of concerete results, from expert systems, to game playing, to automatic translators, to SPAM filters, to name just a few. My own research is in Bioinformatics, where machine learning techniques have proven very valuable, and are constantly being used.
Is this real intelligence? You can argue that it is not. But most AI researchers understand that, for now, achieving human-like intelligence a pipe dream: there are neither the computational tools nor the biological understanding to tackle a problem of this magnitude. Instead, the field is slowly advancing in various directions, making some progress in bridging skill-sets that require human intelligence with those that are best accomplished by computers. That is what AI research is really about: it is real, it is rigorous, and it produces tangible, useful results (rather than some abstract notion of being able to fool humans in a chat session.) Calling it a "fraud" out of ignorance hardly changes matters.
The Turing test is not relevant for one simple reason - it does nothing to connect the lines of research that are currently being actively (and - gasp!- successfully, on occasion) pursued in the field of AI. Yes, fifty years ago AI researchers thought that computers would be able to mimic the behavior of human beings: but the field was just emerging at the time, and people didn't have an understanding of the tremendous complexities involved. (Consider, for comparison, that at the time calculus emerged, people thought that it would be the ultimate tool for explaining all mysteries Life, Universe, and Everything.) Since that time however, these complexities became apparent - many problems in AI have been shown to be Turing-undecidable (no algorithmic solution is possible), and virtually every open problem in the field is at least NP-hard. Furthermore, you cannot really develop human-like intelligence without understanding how the human mind works, and that understanding is sadly lacking. It took hundreds of millions of years for evolution to produce intelligent beings: and this is intelligence is manifested through a massively parallel, largely mysterious mechanism (the brain). Is it even remotely reasonable to expect that researchers would be able to emulate this poorly understood mechanism using vastly different underlying hardware in only 50 years?
The point - one that you are so throughly miss - is that the goal of AI is not to build cute robot friends for the human race, but to devise complex systems to solve specific problems that currently require human intelligence. As such, the field of AI has long ago been split into intersecting sub-domains of machine learning, knowledge represetnation, planning, natural language understanding, e.t.c. Each of those has had some degree of success, and produced plenty of concerete results, from expert systems, to game playing, to automatic translators, to SPAM filters, to name just a few. My own research is in Bioinformatics, where machine learning techniques have proven very valuable, and are constantly being used.
Is this real intelligence? You can argue that it is not. But most AI researchers understand that, for now, achieving human-like intelligence a pipe dream: there are neither the computational tools nor the biological understanding to tackle a problem of this magnitude. Instead, the field is slowly advancing in various directions, making some progress in bridging skill-sets that require human intelligence with those that are best accomplished by computers. That is what AI research is really about: it is real, it is rigorous, and it produces tangible, useful results (rather than some abstract notion of being able to fool humans in a chat session.) Calling it a "fraud" out of ignorance hardly changes matters.
