tag:blogger.com,1999:blog-1145198247762305957.post6566273444815665594..comments2019-06-18T08:13:10.086-05:00Comments on Setting Things Straight: Silas Barta, information theorist by nightSilas Bartahttp://www.blogger.com/profile/09480427306873460464noreply@blogger.comBlogger18125tag:blogger.com,1999:blog-1145198247762305957.post-2839602448726792982010-07-19T19:17:11.917-05:002010-07-19T19:17:11.917-05:00Posting a comment here because Landsburg's pos...Posting a comment here because Landsburg's post is now closed to comments. He substitutes well-ordering for individual well-ordering axioms for each subset of N and then writes: <br /><br />"Okay, our description of the natural numbers just got infinitely long, but at least it’s infinitely long in a simple sort of way. We’ve added an infinite number of axioms, but they all fit the same simple pattern—a pattern that you could easily train your computer to recognize."<br /><br />This misses a major issue. In order for that to be equivalent to well-ordering one needs an uncountable number of axioms (since the cardinality of the powerset of N is the cardinality of the continuum.) If one tries to do this in a computable fashion one ends up with a structure that is strictly weaker than what you get with full well-ordering. I can in fact add to his axiomatic system an axiom to the effect that there's a non-empty subset of N that has no least element, and that will be consistent but omega-inconsistent (I'm glossing over some details here but that's roughly what happens).Joshuahttps://www.blogger.com/profile/00637936588223855248noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-50048264364606932922009-12-16T22:46:04.012-06:002009-12-16T22:46:04.012-06:00What (I think) you are missing is that the incompl...<i>What (I think) you are missing is that the incompleteness of Peano arithmetic implies the existence of nonstandard models, which means that Peano arithmetic is NOT a complete description of arithmetic. Nor does any computable description exist.</i><br /><br />Actually, computable descriptions of addition, subtraction, multiplication, and division, applied to natural numbers, do in fact exist. That <a href="http://www.thebigquestions.com/2009/11/04/what-are-you-surest-of/#comment-206" rel="nofollow">was your definition</a> or arithmetic, remember?<br /><br />Again, you're equating the complexity of X, with the complexity of the task of classifying things built from X. It's the same mistake as if you said, "It takes a lot of data to generate a list of all brick structures that won't fall down; therefore, bricks are complex."<br /><br /><i>Tell you what: Since you are not the only reader who is badly confused about this issue, I'll post a fuller explanation on my own blog tomorrow morning. Then you can let me know if it's still not clear.</i><br /><br />Go for it! But you might want to settle on a definition of complexity before deciding that <i>your</i> views are the non-confused ones. First, you objected that Kolmogorov complexity isn't the only definition of complexity, then your said your usage is consistent with it. And you've added lots of analysis and (re)definition that wasn't in your blog post or book, which can't do anything to substantiate your *original* claims being correct.Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-53545199534270657342009-12-16T16:14:42.727-06:002009-12-16T16:14:42.727-06:00- You say that math is complex because you can'...<i>- You say that math is complex because you can't generate all true statements about arithmetic on a computer.<br /><br />- This only counts as complex because you use a non-standard, counterintuitive, never-used definition of complexity that you never spell out explicity (which would make you see its clumsiness).</i><br /><br />What (I think) you are missing is that the incompleteness of Peano arithmetic implies the existence of nonstandard models, which means that Peano arithmetic is NOT a complete description of arithmetic. Nor does any computable description exist. <br /><br />I'll spell that out again: The fact that the true statements about arithmetic can't be generated by a computer *implies* that there are non-standard models of the axioms, which *means* that no computer can completely describe the natural numbers. <br /><br />The non-existence of a computable description is very much in the spirit of the Kolmogorov definition of complexity. <br /><br />Tell you what: Since you are not the only reader who is badly confused about this issue, I'll post a fuller explanation on my own blog tomorrow morning. Then you can let me know if it's still not clear.Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-81932188861871335972009-12-16T10:23:07.025-06:002009-12-16T10:23:07.025-06:00Yes, and in pretty much the same way that the set ...<i>Yes, and in pretty much the same way that the set of true statements in arithmetic is more complex than the set of true statements in euclidean geometry.</i><br /><br />No, not the same way. The strings I listed are more (or less) complex depending on how long a program it takes to describe them.<br /><br />The two axiom sets that you listed differ in complexity -- by your definition -- because it takes a different number of bits to categorize, into true or false, the entire set of statements that can be generated from those axiom sets. That's *not* "pretty much the same way", and I don't know who speaks of complexity like that.<br /><br />I'd go over the point about geometry vs arithmetic (the problem is not the theorem but your different standards for what counts as a statement within each of them) but that's a relatively minor error compared to the rest of the problems with your claim, which persist even if your definition were valid. To summarize:<br /><br />- You say that math is complex because you can't generate all true statements about arithmetic on a computer.<br /><br />- This only counts as complex because you use a non-standard, counterintuitive, never-used definition of complexity that you never spell out explicity (which would make you see its clumsiness).<br /><br />- You want to import all of the standard implications and connotations about complexity as if you were using the normal definition.<br /><br />Again, <i>that's not how it works</i>. The very reason people use math in the first place is because it's simple, which allows one to efficiently compress descriptions of our observations. If it were actually the most complex "thing in the universe" (and I haven't even touched your Platonism, which is another can of worms), there would be no point to expressing scientific laws in terms of it!<br /><br />If I had to say, I think the word you're looking for is not "complexity" but "usefulness" or "expressiveness".Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-53410694402259251282009-12-15T23:56:39.243-06:002009-12-15T23:56:39.243-06:00It's why people intuitively see 1111111111 as ...<i>It's why people intuitively see 1111111111 as being less complex than 1010101010, which in turn is less complex than 1001011110.</i><br /><br />Yes, and in pretty much the same way that the set of true statements in arithmetic is more complex than the set of true statements in euclidean geometry. <br /><br />The various theorems to this effect (e.g. Tarski's) will not go away just because you happen not to understand them---as is well evidenced by your vision of "lining up squares". You might attempt the exercise of restating the Peano axioms in (say) Tarski's language of geometry if you want to see where your vision fails.Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-17326375132756750782009-12-15T23:26:42.431-06:002009-12-15T23:26:42.431-06:00Re: Kolmogorov complexity. It's more than jus...Re: Kolmogorov complexity. It's more than just an issue of you not using the standard definition of complexity in information theory; it's that that's the appropriate definition to use in the fields you discuss in your book and the blog post. I certainly don't deem it the "only allowable use of the word", but if you're going to deviate, you need to at least define complexity explicitly, which you did in neither your book nor the blog. Heck, you should have defined it even if you *were* just going with Kolmogorov complexity! <br /><br />(Your insinuation I'm trying to force-fit things into a field just because I know a lot about it is noted, petty, and wrong. I do appreciate the promotion to expert though.)<br /><br />But even if I'm wrong about K-complexity's appropriateness for this topic area, your definition *still* doesn't match the everyday use of the term, which actually does follow Kolmogorov complexity quite closely. It's why people intuitively see 1111111111 as being less complex than 1010101010, which in turn is less complex than 1001011110. Nowhere have I have seen a lay conception of "complexity of X" that hinged on how many bits of data you need to generate all true statements that involve permutations of components of X, as you seem to be defining the term. And such an esoteric definition of complexity takes most of the significance away from the conclusions predicated on it.<br /><br />Yes, if what you were saying about *normal* complexity were true, that would be cool, but you're only saying it about Landsburg-complexity, which no one cares about.<br /><br />(Incidentally, I've been giving your comparison of Euclidean geometry less attention, but that's just as wrong too -- you're placing a tighter constraint on what counts as a geometric statement compared to an arithmetic one. After all, you can line up squares and make statements about them that are isomorphic to statements about the natural numbers.)<br /><br /><i>"genome" is a term of art in genetics, but it has a respectable history of everyday use as "complete set of instructions for assembling proteins into a living being", which includes not just a sequence of symbols but an attribution of meaning to each of those symbols, with the meanings incorporating the way the proteins fit together, etc etc.</i><br /><br />Not any place I'm aware of. Rather, people assumed that the literal genome would get them the benefits of the figurative genome, and the confusion has persisted despite this being proved wrong. Whatever history that is, isn't respectable. And you're not doing anyone a favor by perpetuating this misconception. <br /><br /><i>Incidentally, you used the word "produce" in one of your comments. Are you aware that "produce" is a term of art in the field of economics and that your use of that word does not coincide with the standard economic definition at all? Would it strike you as childish if I used that as a pretext to dismiss everything of content in your comments?</i><br /><br />It strikes me as childish for you to make this comparison. If I were dealing with subject matter where "produce" in the economic sense is the standard definition, but used it in a completely different way, foreign to that subject matter and the lay sense, AND made an argument crucially dependent on the meaning of "produce", AND lacked a definition anywhere, AND which made my thesis trivial if my definition actually *were* used, then yes, all of the claims I've made about your use of "complexity" would be justified.Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-48970226759964428882009-12-15T22:21:57.651-06:002009-12-15T22:21:57.651-06:00"Complexity" is already taken -- it mean...<i>"Complexity" is already taken -- it means something else.</i> "Complexity" means many different things in many different contexts, some of those meanings being precise and formal and others being colloquial. The fact that it's a term of art in information theory does not detract from its usefulness as an everyday English word. If your complaint is that I didn't use the word "complexity" in the way it's used in the field in which you happen to be an expert, then I plead guilty. The fact remains that arithmetic, in the everyday sense of the word, is enormously more complex than, say, euclidean geometry, in the sense that all true statements of euclidean geometry can be generated from finitely many bits of data and there is a sense in which "most" true statements of arithmetic cannot. Come to think of it, that's pretty damn close in spirit to the technical Kolmogorov definition of complexity that you seem to think is the only allowable use of the word. <br /><br />Re the genome, again---"genome" is a term of art in genetics, but it has a respectable history of everyday use as "complete set of instructions for assembling proteins into a living being", which includes not just a sequence of symbols but an attribution of meaning to each of those symbols, with the meanings incorporating the way the proteins fit together, etc etc. And what matters to my argument that the genome, in this broader more colloquial sense, is encodable as a fragment of arithmetic. <br /><br />Incidentally, you used the word "produce" in one of your comments. Are you aware that "produce" is a term of art in the field of economics and that your use of that word does not coincide with the standard economic definition <i>at all</i>? Would it strike you as childish if I used that as a pretext to dismiss everything of content in your comments?Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-89287876773088826652009-12-15T19:07:51.235-06:002009-12-15T19:07:51.235-06:00By "all true statements about arithmetic"...<i>By "all true statements about arithmetic", I meant "all true statements in the language of arithmetic". ... I am reasonably certain that when I refer to "statements about arithmetic", virtually all readers will interpret that as intended, e.g. statements like "two plus two equals four" and "every number is the sum of (at most) four squares". ... No computer program can generate all true statements in the language of arithmetic. </i><br /><br />That still doesn't help you. Computer programs lack this ability because of the problem of distinguishing true from false statements. What definition of complexity hinges on "how difficult it is to make determinations about things your produce with it"?<br /><br /><i>One cannot say the same for euclidean geometry. This is a substantive and relevant distinction.</i><br /><br />Then make the distinction using the appropriate terms. "Complexity" is already taken -- it means something else.<br /><br /><i>Again, you are quibbling about my informal use of language in an informal venue. Whether the genome is sufficient to "encode" life depends on what you mean by "encode".</i><br /><br />No, it doesn't. No matter what definition of "encode" you use, you still need more than the genome in order to make sense of the base-4 string.<br /><br /><i>Again, I'd have thought the meaning was clear from context, and whether or not you like my choice of language, the fundamental point remains that all of the A's, C's, G's, T's, and the chemical reactions in which they take part, etc., etc., are all imbeddable in a finite fragment of arithmetic.</i><br /><br />Your text (and discussion in your book) don't show an understanding of this distinction. Why bring up the genome at all, if you understand that it's not the same thing as human life?Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-29855254063049040572009-12-15T18:25:59.318-06:002009-12-15T18:25:59.318-06:00...you can use just a small part of [the natural n...<i>...you can use just a small part of [the natural numbers] to encode the entire human genome. That makes the natural numbers more complex than human life.</i><br /><br />You've permitted yourself the luxury of letting sophisticated readers infer the specific content of your informal statements about arithmetic by invoking Godel. You ought to permit your sophisticated readers the luxury of taking you to task when you write something that strongly and falsely implies that the human genome specifies human life without quibbling yourself.Cyannoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-4835447289897741492009-12-15T17:58:59.980-06:002009-12-15T17:58:59.980-06:00"No computer can generate all true statements...<i><br />"No computer can generate all true statements about arithmetic."</i><br /><br />Okay, I see your issue. By "all true statements about arithmetic", I meant "all true statements in the language of arithmetic". The language is imprecise, but I think appropriately so given that I wasn't writing a mathematics textbook. I am reasonably certain that when I refer to "statements about arithmetic", virtually all readers will interpret that as intended, e.g. statements like "two plus two equals four" and "every number is the sum of (at most) four squares". For a more sophisticated reader, I'd have expected this to be clear from context, given my invocation of Godel.<br /><br />But since this wasn't clear to you, let me clarify it now: No computer program can generate all true statements in the language of arithmetic. One cannot say the same for euclidean geometry. This is a substantive and relevant distinction.<br /><br /><i>Do you now agree that encoding the entire human genome is insufficient to encode human life, even without resorting to the concept of a soul? </i> <br /><br />Again, you are quibbling about my informal use of language in an informal venue. Whether the genome is sufficient to "encode" life depends on what you mean by "encode". Again, I'd have thought the meaning was clear from context, and whether or not you like my choice of language, the fundamental point remains that all of the A's, C's, G's, T's, and the chemical reactions in which they take part, etc., etc., are all imbeddable in a finite fragment of arithmetic.Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-38746986278943836892009-12-15T17:33:33.374-06:002009-12-15T17:33:33.374-06:00By the way, Steven_E._Landsburg: Do you now agree ...By the way, Steven_E._Landsburg: Do you now agree that encoding the entire human genome is insufficient to encode human life, <i>even</i> without resorting to the concept of a soul? Because at the very least, you need to list a large number of chemical reactions.Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-59123610817371303172009-12-15T17:20:14.994-06:002009-12-15T17:20:14.994-06:00No, I wouldn't be wrong, and you've shown ...No, I wouldn't be wrong, and you've shown no distinction. Remember, you've decided to count any statement <i>about</i> X towards X's complexity. That was your justification for calling arithmetic (or math, or whatever) complex:<br /><br />"No computer can generate all true statements about arithmetic."<br /><br />Well, since you can just involve arbitrary arithmetic predicates in statements <i>about</i> apples, or Euclidean geometry, or whatever else, then, by that metric, <i>everything</i> is maximally complex.<br /><br />But of course, no one else considers "the size of the set of true statements about X" to be the definition of X's complexity.<br /><br />What you seem to want to do is use your own definition of complexity, involving how many statements can be generated about something, in one context, and then abandon it in others.Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-90690105311264839772009-12-15T17:09:36.647-06:002009-12-15T17:09:36.647-06:00you can point to anything and say "no compute...<i>you can point to anything and say "no computer can generate all true statements about this".</i><br /><br />Well of course you can <i>say</i> this, but only if you're willing to be <i>wrong</i>. You would be wrong, for example, if you said it about euclidean geometry or the theory of real closed fields. But you'd be right if you said it about arithmetic. That seems like kind of an important distinction.Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-41224098862710680542009-12-15T15:32:21.578-06:002009-12-15T15:32:21.578-06:00No computer can generate all true statements about...<i>No computer can generate all true statements about arithmetic; arithmetic is too complex for that.</i><br /><br />That doesn't make arithmetic (or "the laws of arithmetic") complex. That means you can produce complex things within the system of arithmetic. Again, if you include that as part of the definition, then *everything* counts as complex, and the term has no meaning -- you can point to anything and say "no computer can generate all true statements about this".Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-65787061852750634022009-12-15T13:39:53.664-06:002009-12-15T13:39:53.664-06:00Arithmetic can be simulated on a computer; proofs ...<i>Arithmetic can be simulated on a computer; proofs for all true statements in math (by the theorems you gave) cannot.</i><br /><br />The theorems I gave are not about math; they are about arithmetic. No computer can generate all true statements about arithmetic; arithmetic is too complex for that.Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-7077151353303381892009-12-15T13:38:25.911-06:002009-12-15T13:38:25.911-06:00This comment has been removed by a blog administrator.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-11768260597825146722009-12-15T12:05:27.609-06:002009-12-15T12:05:27.609-06:00@Steven_E._Landsburg: Thank you for your comment.
...@Steven_E._Landsburg: Thank you for your comment.<br /><br />However, you're still equivocating: between the building blocks, and what you can do with the building blocks. (And, again, between math in general and arithmetic, which you took pains to distinguish before, though I can't find the comment on your blog anymore.)<br /><br />Arithmetic <i>can</i> be simulated on a computer; proofs for all true statements in <i>math</i> (by the theorems you gave) cannot. But the complexity of that proof set is not evidence for the complexity of arithmetic (or math, whichever meaning you want to use this time), any more than a complex structure made of bricks is evidence that the bricks are complex. You can't get there from here.<br /><br />Your original claim as that the natural numbers, combined with the laws of arithmetic, are more complex than human life. But you can write a simple program that specifies the basic arithmetic operations and generates all natural numbers.<br /><br />If you're going to count the complexity of the product of the tools toward the tools themselves, then math is not special in being complex -- in fact, nothing is, since by that premise, everything is complex, and the term has no meaning.Silas Bartahttps://www.blogger.com/profile/09480427306873460464noreply@blogger.comtag:blogger.com,1999:blog-1145198247762305957.post-45447538778320936322009-12-14T22:05:13.910-06:002009-12-14T22:05:13.910-06:00Don't worry -- the process terminates: once yo...<i>Don't worry -- the process terminates: once you've described the generative model that puts all of these concepts together in a way that yields a description of human life as its output.</i><br /><br />Yes, it terminates in a combinatorial structure, which is to say that it terminates in a fragment of arithmetic. <br /><br /><br />Indeed, it terminates in a particularly simple fragment of arithmetic---one that can be simulated (in principle) on a computer, which is decidedly not true of arithmetic as a whole. (This is essentially the content of Godel's first incompleteness theorem, or, better, Tarski's theorem on the undefinability of truth.)Steven E. Landsburghttp://www.thebigquestions.com/blognoreply@blogger.com