On Thu, Jul 4, 2013 at 7:21 AM, James Bowery <[email protected]> wrote:
Well since we're talking measurement and theory in the natural sciences, > one is operating on nature and one does have a model of nature which is > formal in the sense that any theory is formal. > I think we are largely in agreement here. There are perhaps two or three different "formal" approaches that are possible -- there's the formality of a formal definition, i.e., "intelligence is A and B," where you can rigorously show that A and B are satisfied or not, in a mathematical sense. And then there's the formality of a procedure -- "its not clear exactly what intelligence is and whether computers can have it, but we think we can rigorously detect some examples of intelligence being used that could potentially overlap with what computers can do now or in the future. For our experiment, we'll try to place bounds the question by doing C and D, and whatever we find, it will be interesting and statistically sound." And then there's the formality of a model -- "we don't know exactly what intelligence is or whether computers can have it, but we need to approach the problem systematically and relate the results to other experiments, so here are our general assumptions: E and F." It would probably be difficult to keep these three dimensions apart in actual experiments. But it seems to me that the first kind of formality could lead people into to assuming the answer implicitly in the question; for example, "intelligence is the ability to solve a certain class of NP-hard problems together with <fill in three other abilities>." Eric
