Back in 1972 the Philosopher Hubert Dreyfus wrote an interesting book called "What Computers Can't Do" as a critique of the position taken by the AI gurus of the day such as Marvin Minski.
Minski had been making statements such as in 1970, "Once the computers got control, we might never get it back. We would survive at their sufferance. If we're lucky, they might decide to keep us as pets." Hmm well 46 years later and there's not much sign of that.
Just to rub it in, Dreyfus followed up with "What Computers Still Can't Do" in 1992
Dreyfus was greatly influenced by the later Wittgenstein, arguing that rule based symbol systems could never produce human like intelligence because they would always lack context. So for example the sentence "The bill is large" has quite different meaning depending on whether you're in a restaurant looking at a slip of paper demanding an excessive payment or in a zoo looking at a large duck.
Rule based inference engines struggle with this type of ambiguity which is why modern AI systems tend to be effective only in relatively limited domains and the success at the Turing test is as far away as its ever been. It's true that some researchers have claimed sucess but the tests were of such a short period, 5 mins, that they were hardly credible.
Dreyfus questions that intelligence really is based on rules. He discusses Socrates who would stroll around Athens picking on unwary passers by and asking them one of his famous Socratic questions such as "What is courage". They would inevitably start giving examples of courage and Socrates would reply "no you fool, i dont want examples, i want the rules by which you judge an action as courageous". Dreyfus wonders if actually Socrates was wrong and intelligence is nothing to do with rules and actually about recognising a problem as similar to examples in the past that were relevant to solving the current problem and extrapolating from them. If true then intelligence would be more to do with memory and creative imagination than rules. Whether a machine that could store and retrieve the vast complexity of memory required for human like intelligence, could ever be built, is an interesting question. Certainly attempts to pin down human memory aren't going that well.
Minski had been making statements such as in 1970, "Once the computers got control, we might never get it back. We would survive at their sufferance. If we're lucky, they might decide to keep us as pets." Hmm well 46 years later and there's not much sign of that.
Just to rub it in, Dreyfus followed up with "What Computers Still Can't Do" in 1992
Dreyfus was greatly influenced by the later Wittgenstein, arguing that rule based symbol systems could never produce human like intelligence because they would always lack context. So for example the sentence "The bill is large" has quite different meaning depending on whether you're in a restaurant looking at a slip of paper demanding an excessive payment or in a zoo looking at a large duck.
Rule based inference engines struggle with this type of ambiguity which is why modern AI systems tend to be effective only in relatively limited domains and the success at the Turing test is as far away as its ever been. It's true that some researchers have claimed sucess but the tests were of such a short period, 5 mins, that they were hardly credible.
Dreyfus questions that intelligence really is based on rules. He discusses Socrates who would stroll around Athens picking on unwary passers by and asking them one of his famous Socratic questions such as "What is courage". They would inevitably start giving examples of courage and Socrates would reply "no you fool, i dont want examples, i want the rules by which you judge an action as courageous". Dreyfus wonders if actually Socrates was wrong and intelligence is nothing to do with rules and actually about recognising a problem as similar to examples in the past that were relevant to solving the current problem and extrapolating from them. If true then intelligence would be more to do with memory and creative imagination than rules. Whether a machine that could store and retrieve the vast complexity of memory required for human like intelligence, could ever be built, is an interesting question. Certainly attempts to pin down human memory aren't going that well.