> So, the question then is, how to APPLY that to a general set of problems. That's the tricky thing. As far as we know, IQ is the best proxy for general probl

> So, the question then is, how to APPLY that to a general set of problems. That's the tricky thing. As far as we know, IQ is the best proxy for general problem solving ability. From what I've read/heard from intelligence researcher; however, is that while you can train folks to excel at certain types of problems, it doesn't really improve one's general problem solving ability. Think of an IQ test as a random sampling of problem solving exercises of various types from the domain of all possible problem solving exercises. Now imagine generating a whole bunch of such tests, and then using those tests on a bunch of people to setup your distribution (i.e. set the IQ scoring for each test). While it is true that in general the scores are going to be highly correlated across tests, each test is going to have a different composition in terms of the kinds of problems it presents. From what I understand, training may help on certain classes of problems (and thus on particular tests) but there isn't any training that will meaningfully improve the average of all your scores across all of the tests.