His essay has the right conclusion, but the premises are mostly false. Some people have 160 IQ (we can't generally know who, bragging about having them is generally a sign of a problem), historical estimate have some issues (all do), intelligence does mostly linearly relate to success and often superlinearily in the right tail, measureme…
His essay has the right conclusion, but the premises are mostly false. Some people have 160 IQ (we can't generally know who, bragging about having them is generally a sign of a problem), historical estimate have some issues (all do), intelligence does mostly linearly relate to success and often superlinearily in the right tail, measurement error goes up away from the mean using fixed-length tests (but we can just use other tests). Not one of his best essays.
I think you're confusing the map with the territory. IQ isn't intelligence; it's a *measure* of intelligence. So to say, "Some people have 160 IQ (we can't generally know who," doesn't really make sense. We can say, "we can be confident that there are some people who, if tested, would score 160," but the whole point here is that that's meaningless.
IQ is the metric, not the measure. IQ is just a shorthand for writing about age relative differences (115 = 1 standard deviation above age peers).
There is no confusion. If we had a perfect IQ test (standard error of measurement = 0 for entire range), there are some people who would attain a score of 160.
We have a very comprehensive program aimed squarely at producing intelligent behavior. It relies 0.0% on any kind of IQ research, because the IQ field has essentially nothing useful to say about the subject of its own investigation. Why are some people good at IQ tests? What do they do to excel at them? How could those techniques inform the AI program? If we ever learn it won't be thanks to anyone who "studies" IQ.
His essay has the right conclusion, but the premises are mostly false. Some people have 160 IQ (we can't generally know who, bragging about having them is generally a sign of a problem), historical estimate have some issues (all do), intelligence does mostly linearly relate to success and often superlinearily in the right tail, measurement error goes up away from the mean using fixed-length tests (but we can just use other tests). Not one of his best essays.
I think you're confusing the map with the territory. IQ isn't intelligence; it's a *measure* of intelligence. So to say, "Some people have 160 IQ (we can't generally know who," doesn't really make sense. We can say, "we can be confident that there are some people who, if tested, would score 160," but the whole point here is that that's meaningless.
IQ is the metric, not the measure. IQ is just a shorthand for writing about age relative differences (115 = 1 standard deviation above age peers).
There is no confusion. If we had a perfect IQ test (standard error of measurement = 0 for entire range), there are some people who would attain a score of 160.
We have a very comprehensive program aimed squarely at producing intelligent behavior. It relies 0.0% on any kind of IQ research, because the IQ field has essentially nothing useful to say about the subject of its own investigation. Why are some people good at IQ tests? What do they do to excel at them? How could those techniques inform the AI program? If we ever learn it won't be thanks to anyone who "studies" IQ.