What I don’t understand is so much insistence that Occam’s Razor applies only to explanations you address to God. Or else how do you avoid the observation that the simplicity of an explanation is a function of whom you are explaining to ? In the post, you actually touch on the issue, only to observe that there are difficulties interpreting Occam’s Razor in the frame of explaining things to humans (in their own natural language), so let’s transpose to a situation where humans are completely removed from the picture. Curiously enough, where the same issue occurs in the context of machine languages it is quickly “solved”. Makes one wonder what Occam—who had no access to Turing machines—himself had in might.
Also, if you deal in practice with shortening code length of actual programs, at some point you have exploited all the low lying fruit; further progress can come after a moment of contemplation made you observe that distinct paths of control through the code have “something in common” that you may try to enhance to the point where you can factor it out. This “enhancing” follows from the quest for minimal “complexity”, but it drives you to do locally, on the code, just the contrary of what you did during the “low-lying fruit” phase, you “complexify” rather than “simplify” two distinct areas of the code to make them resemble each other (and the target result emerges during the process, fun). What I mean to say, I guess, is that even the frame proposed by Chaitin-Kolmogorov complexity gives only fake reasons to neglect bias (from shared background or the equivalent).
What I don’t understand is so much insistence that Occam’s Razor applies only to explanations you address to God. Or else how do you avoid the observation that the simplicity of an explanation is a function of whom you are explaining to ? In the post, you actually touch on the issue, only to observe that there are difficulties interpreting Occam’s Razor in the frame of explaining things to humans (in their own natural language), so let’s transpose to a situation where humans are completely removed from the picture. Curiously enough, where the same issue occurs in the context of machine languages it is quickly “solved”. Makes one wonder what Occam—who had no access to Turing machines—himself had in might.
Also, if you deal in practice with shortening code length of actual programs, at some point you have exploited all the low lying fruit; further progress can come after a moment of contemplation made you observe that distinct paths of control through the code have “something in common” that you may try to enhance to the point where you can factor it out. This “enhancing” follows from the quest for minimal “complexity”, but it drives you to do locally, on the code, just the contrary of what you did during the “low-lying fruit” phase, you “complexify” rather than “simplify” two distinct areas of the code to make them resemble each other (and the target result emerges during the process, fun). What I mean to say, I guess, is that even the frame proposed by Chaitin-Kolmogorov complexity gives only fake reasons to neglect bias (from shared background or the equivalent).