Probably the best place to start with the notion of artificially intelligent machines, is to get some perspective on exactly what comprises intelligence, or other similar concepts like awareness or se
Probably the best place to start with the notion of artificially intelligent machines, is to get some perspective on exactly what comprises intelligence, or other similar concepts like awareness or sentience and so forth. Although it might seem an awfully long time ago now, the debate on the nature of intelligence began back in the mid-nineteenth century, in a correspondence between the inventor of the first mechanical computer - Charles Babbage - and the first programmer - Lady Ada Lovelace. Lovelace insisted that The analytical engine has no pretensions to originate anything. It can do whatever we know how to order it to perform; everything a computer might ever do was dictated by its programming and nothing new might emerge. Doubtless the obvious mechanical nature of Babbages Analytical Engine made the idea that it might be somehow equivalent to a person quite insulting. These days, however, computer systems do perform tasks widely regarded as requiring original thought - the derivation of scientific theories from experimental data, the design of new materials, and so on. These systems are not the sort of thing you might converse with, of course, and this begs the question What is originality anyway? Do we, in fact, always act with originality? Or do we just copy patterns a lot of the time?. It seems that while the ability to produce new, original material is a requirement of an intelligent system, that, in and of itself, is not enough.
In the 1940s, a British mathematician named Alan Turing set the mathematical world on its ear when he provided a rigorous theoretical grounding of the workings of a universal information processing machine - the programmable computer. Whilst Babbage had been the first to design a programmable computer, Turing, almost a century later, provided the theories that demonstrated the extraordinary capabilities of the concept. What Turing demonstrated, more or less, was that a computer, correctly programmed, could emulate the behavior of any of an infinite number of information processing machines - machines that performed a well defined, step by step task that involved the manipulation of symbols. So the question then arose, are the workings of the human mind a well-defined step-by-step task? Are they something a computer, as defined by Turing, and as used today, might perform?
Rather than actually attempting to define the exact nature of human intelligence - quite a task - Turing sidestepped the question raised by his work. His novel, and these days much criticised approach to deciding if the workings of the human mind can be replicated by a computer, is known as the Turing Test.
Whilst the Turing Test has, and does continue to focus AI research on fruitful areas - natural language processing and knowledge representation, for instance - there still seems to be something missing. The Turing test is not generally regarded as a hard criteria for machine intelligence these days, or even as evidence that the processes in the human mind can be performed by a computer. Most of the objections to the test are based upon the possibility of performing a task without any awareness of the content of the task at hand; reading another language by just pronouncing the sounds, for example. We all have rich internal emotions, thoughts, memories and so forth - just passing the Turing test doesnt guarantee that a machine has a similar experience.
Not only is such a rich internal life important to the acknowledgment of actual intelligence, at least as far as most people seem concerned, it also serves to bring into focus one of the more pure motivations for work in AI: attempting to better understand ourselves, and what it is to be human. You have to wonder, of course, if a machine could hold an emotional conversation and have no emotion behind it; surely a person could too? Weve all had conversations where the other person wasnt quite with it, but how can we be sure that anyone else is, in fact, there at all? How do we know someone is really a person, and not just a convincing automaton? Is there some spark that might convince us, and if so, can we fit it into a machine?