The issue with AI has nothing to do with ignorance or arrogance. The basic problem is that intelligence can’t be meaningfully defined or meaningfully quantified. Documented fact: Richard Feynman had a measured I.Q. of 120. Documented fact: Marilyn Vos Savant had a measured I.Q. of 180 or 200, depending on which test you place more faith in. Documented fact: Feynman made a huge breakthrough in physics, Vos Savant has accomplished nothing worth mentioning in her life. I.Q. measurements fail to measure intelligence in any meaningful way.
Here’s another fact for you. Louis Terman collected a group of so-called “geniuses” sieved by their high I.Q. scores. Two future nobel prize winners, Shockley and Alvarez, got tested but discarded by Terman’s I.Q. tests and weren’t part of the group.
Question: What does this tell you about current methods for measuring intelligence?
There is no evidence that people can meaningfully define or objectively measure intelligence. Rule of thumb: if you can’t define it and you can’t measure it objectively, you can’t do science about it.
[Remainder of gigantic comment truncated by editor.]
I’m surprised nobody brought this up at the time, but it’s telling that you’ve only picked out examples of humans when discussing intelligence, not bacteria or rocks or the color blue. I submit that the property is not as unknowable as you would suggest.
Nobel prizes aren’t based only on intelligence, but also on drive, persistence and also a little luck (mainly the luck to find something interesting to work on that nobody else has yet solved).
After all “1% inspiration and 99% perspiration” yes?
Drive and persistence are part of intelligence, at least in the sense that any useful AI would have to have them. Saying it measures luck is just saying that it’s imprecise.
That said, it’s not going to measure all the different components of intelligence in the way we want.
The problem isn’t that it can’t be meaningfully defined or quantified. The problem is that it hasn’t been. I have no idea how hard it is to do that. It may very well be beyond anything any human can do, but it’s theoretically possible.
In the hypothetic universe, addition certainly could be defined, it’s just that nobody in that universe knew how.
Intelligence is a multidimensional concept that is not amenable to any single definition or quantization. Take for instance the idea of “the size of a tree.” Size could mean height, drip radius, mass, volume of smallest convex polyhedron that contains the whole organism, volume of water displaced if the tree was immersed in a tank, trunk girth at 6 feet, etc. The tallest redwood is taller than the tallest sequoia, but isn’t the sequoia bigger? Why is it bigger? Because it has greater mass? But what of the biggest banyan? It has a greater mass than both the redwood and the sequoia.
The problem with intelligence is not that it’s not quantifiable, but that different researchers use different mapping functions all the while pretending they’re measuring the exact same thing, heaping up the confusion. If you pick one specific mental activity (arithmetic, visual memory, music-compositional ability, language processing), it is rarely very difficult to measure and rank people by their adeptness. If, on the other hand, you try to come up with a “good” way to map many different intelligences together onto some scale, you’re going to be terrible at using this scale to predict individual performance at specific tasks. Further, individuals with low IQ (or other attempted measure at general intelligence) may be brilliant at specific tasks because of their low IQ in that because much of their brain is dedicated to that task, they have little left over for anything else. This is especially true of many autistic individuals.
In the end, intelligence is rather easy to define if you recognize it as the multifaceted phenomena that it is.
The issue with AI has nothing to do with ignorance or arrogance. The basic problem is that intelligence can’t be meaningfully defined or meaningfully quantified. Documented fact: Richard Feynman had a measured I.Q. of 120. Documented fact: Marilyn Vos Savant had a measured I.Q. of 180 or 200, depending on which test you place more faith in. Documented fact: Feynman made a huge breakthrough in physics, Vos Savant has accomplished nothing worth mentioning in her life. I.Q. measurements fail to measure intelligence in any meaningful way.
Here’s another fact for you. Louis Terman collected a group of so-called “geniuses” sieved by their high I.Q. scores. Two future nobel prize winners, Shockley and Alvarez, got tested but discarded by Terman’s I.Q. tests and weren’t part of the group.
Question: What does this tell you about current methods for measuring intelligence?
There is no evidence that people can meaningfully define or objectively measure intelligence. Rule of thumb: if you can’t define it and you can’t measure it objectively, you can’t do science about it.
[Remainder of gigantic comment truncated by editor.]
I’m surprised nobody brought this up at the time, but it’s telling that you’ve only picked out examples of humans when discussing intelligence, not bacteria or rocks or the color blue. I submit that the property is not as unknowable as you would suggest.
Nobel prizes aren’t based only on intelligence, but also on drive, persistence and also a little luck (mainly the luck to find something interesting to work on that nobody else has yet solved).
After all “1% inspiration and 99% perspiration” yes?
Drive and persistence are part of intelligence, at least in the sense that any useful AI would have to have them. Saying it measures luck is just saying that it’s imprecise.
That said, it’s not going to measure all the different components of intelligence in the way we want.
Nobel prizes are measuring something (or, more likely, a bunch of things), but is it a good match for what we mean by intelligence?
The problem isn’t that it can’t be meaningfully defined or quantified. The problem is that it hasn’t been. I have no idea how hard it is to do that. It may very well be beyond anything any human can do, but it’s theoretically possible.
In the hypothetic universe, addition certainly could be defined, it’s just that nobody in that universe knew how.
Intelligence is a multidimensional concept that is not amenable to any single definition or quantization. Take for instance the idea of “the size of a tree.” Size could mean height, drip radius, mass, volume of smallest convex polyhedron that contains the whole organism, volume of water displaced if the tree was immersed in a tank, trunk girth at 6 feet, etc. The tallest redwood is taller than the tallest sequoia, but isn’t the sequoia bigger? Why is it bigger? Because it has greater mass? But what of the biggest banyan? It has a greater mass than both the redwood and the sequoia.
The problem with intelligence is not that it’s not quantifiable, but that different researchers use different mapping functions all the while pretending they’re measuring the exact same thing, heaping up the confusion. If you pick one specific mental activity (arithmetic, visual memory, music-compositional ability, language processing), it is rarely very difficult to measure and rank people by their adeptness. If, on the other hand, you try to come up with a “good” way to map many different intelligences together onto some scale, you’re going to be terrible at using this scale to predict individual performance at specific tasks. Further, individuals with low IQ (or other attempted measure at general intelligence) may be brilliant at specific tasks because of their low IQ in that because much of their brain is dedicated to that task, they have little left over for anything else. This is especially true of many autistic individuals.
In the end, intelligence is rather easy to define if you recognize it as the multifaceted phenomena that it is.