Just a note that I have put up a new post at Panda’s Thumb in response to a post by Eric Holloway at the Discovery Institute’s new blog Mind Matters. Holloway declares that critics have totally failed to refute William Dembski’s use of Complex Specified Information to diagnose Design. At PT, I argue in detail that this is an exactly backwards reading of the outcome of the argument.
Commenters can post there, or here — I will try to keep track of both.
There has been a discussion of Holloway’s argument by Holloway and others at Uncommon Descent as well (links in the PT post). gpuccio also comments there trying to get someone to call my attention to an argument about Complex Functional Information that gpuccio made in the discussion of that earlier. I will try to post a response on that here soon, separate from this thread.
Hi Joe
Eric has been talking about the law of information of non growth extensively at peaceful science. He feels compelled by the argument and I don’t really understand it. I hope this discussion surfaces the arguments strengths and weaknesses. If he is right then natural selection is not able to create any new FI as according to the law any combination of randomness and determinism cannot generate FI.
Yes, I noticed the discussions at PS and at UD. He seems very convinced, but I am fairly sure he is wrong about that (unless he redefines FI from what Szostak and Hazen intended). I intend to put up a post here (a couple of weeks hence) asking what is the logic of this, and also how the algorithmic specified complexity criterion makes any sense in a biological situation. I am fairly sure that it doesn’t.
In any case those are both separate from the issue of why Holloway thinks Dembski’s argument has not been demolished. That is what is under discussion in this thread.
colewd,
Bill,
When the dust of the discussion settles down, it’s going to boil down to whose definition of FI is better…
Both sides miss the bigger picture though; the law of conservation of quantum information. I have mentioned it more than once at TSZ and PS.
In embryo development or even in the process of protein synthesis there is a loss of (classical) information. So how the hell can natural selection create new FI if there are losses on the way?
Was this post some kind of typing error? Did you mean to write:
Joe Felsenstein Needs Our Help to Explain Why Dembski’s Argument HAS Been Demolished?
How is he supposed to explain why it hasn’t been?
Maybe I should start a new post: Evolutionists Need Help Explaining Why Their Theory About Some Rocks Being More Smooth Than Others Hasn’t Already Been Demolished.
Or perhaps: Joe Felsenstein Needs Help Proving He Wasn’t Abducted By Aliens.
Creationist commentators, possibly, will have trouble posting at Pandas Thumb. if i’m a sample I’m corralled to one post and thats the end of the discussion. so creationists ID/YEC, would have trouble unless they are more kosher then me.
TSZ is probably better. Free speech and all that.
Anyways.
Well if Dembski’s idea has been or not been refuted in a settled way then a summery of either view can easily be made.
its a little off broadway for me but Dembski always has great ideas and thats why he is well known by those interested in innovative discoveries in science. Agree or don’t agree.
Its a worthy discussion.
First phoodoo needs to explain why the arguments made in the Panda’s Thumb OP are wrong. The links there lead to the papers and blog posts that make those arguments. Have at it — present your argument. Mere assertion won’t do, of course.
Joe Felsenstein,
One less ‘scientific’ reason is because Holloway thinks he’s part of a ‘movement’ based around the concept duo of ‘intelligent’ + ‘design,’ with its chief agitation port of call in Seattle, Washington, USA. Hollaway writes like someone defending a ‘home base’ that can be summarised in among the simplest of ways for self-rationalisation. His ideological IDist approach can really only potentially appeal or make sense to people who are a combination of being, 1) already theists (most likely confused evangelical protestants), 2) suspicious of science, and 3) poorly read in philosophy.
As we know, Dembski once claimed to have ‘retired’ from IDT = Intelligent Design theory (properly capitalised, against the politicised wishes of the author), yet now recently has appeared again in the form of the Walter Bradley Centre, where Holloway is likewise affiliated. So, there’s a post-retirement, re-engagement between Dembski & the DI that might be worth looking into here at TSZ (given that this site grew largely out of opposition to UD, which Dembski started, then abandoned, not unlike the philosophistic owner of this site).
Are natural & artificial intelligences not a topic of interest to many on this channel?
One might wonder when Dembski, Meyer, West, Chapman & the good ole’ folks in Seattle are eventually going to come around on their honour to actually face the ever-present & continued strongest challenge to their now institutionalised ideology of IDism, that which is conveyed to them regularly by Abrahamic monotheists, including scientists, philosophers & theologians who reject IDism & what the IDM is trying to do to biology and ‘evolutionary science’?
Gregory,
Actually, while they weren’t looking I sneaked into Fortress Seattle, some time ago.
The Bradley Center does not interest me. These folks are very concerned to establish that
1. Evolution needs ID (this part does interest me)
2. Only humans can be intelligent, so machines can’t
3. Beings on other worlds can’t be intelligent
4, Multiverses cannot exist because then someone else in one of them might be intelligent.
All of these for basically theological reasons.
At UD, Denyse O’Leary campaigns on a large number of fronts like that. It was a wonder to behold her trying to establish that the Higgs Boson did not exist. She insisted on that because someone had called it “the God particle”, and a particle cannot be God. But they only did that because they were having trouble finding it and someone started calling it “the goddamn particle” and then they had to clean that phrase up. They forgot to tell O’Leary that that was the only reason it was a “God particle”.
As for the Bradley Center, the folks ensconced there don’t know much about AI, so that outfit doesn’t interest me.
Evidently you were writing here as I was writing at PT. I should mention for the sake of others that you and I have had no side discussion on this matter.
Bob Marks is highly competent in computational intelligence. I think it was 1988 when I first read a paper of his on neural nets. He and his colleagues at the University of Washington — I think you’ve heard of the place — did excellent work. But here’s a highly telling anecdote: Back in the mid-90s, artificial intelligence was part of the turf of the IEEE Computer Society. So Bob, as the president of the IEEE Neural Networks Council (sort of a provisional organization, not belonging to any one of the existing societies within the IEEE), pushed an entirely artificial distinction of artificial and computational intelligence. I groaned when I first saw it. (However, back in those days, I assumed that Bob was a pretty good guy. When he started working with Dembski, I continued to assume that he was a pretty good guy. All of the disapprobation he gets from me now, he has gone below and beyond to earn.) Everything that Bob identified as computational intelligence, and not artificial intelligence, was addressed extensively in publications of the traditional AI community. Everything. As I came to know IEEE members who latched onto “computational intelligence” I found that they were, though very bright folks, largely ignorant of the field of artificial intelligence. Anyway, Bob Marks was joined by others in pushing the false and needless AI-CI distinction, and the ultimate outcome was the creation of a new society within the IEEE, the Computational Intelligence Society. So what’s wrong with that? It splintered a field that was not all that large. It kept people who should have been going to the same conferences from knowing about one another’s work. The one good thing I can say about the division is that it led to more journal pages in a booming field that desperately needed them.
So what does this tell us about Bob Marks? He’s long been a politician. He’ll do unsavory stuff to achieve ends that he thinks are worthy. It’s not terribly surprising that he, though an engineer, is now heading an operation dedicated to the generation of conservative religio-political propaganda.
However, I must say, as I did at PT, that when Bob makes technical claims about machine intelligence, he usually gets them right. Eric Holloway and Jonathan Bartlett commonly get them very wrong. Evidently the editorial operation is being run by O’Leary, not Marks. I don’t think much of Bob attaching his name to the operation, and then ignoring quality control. But this is, relative to what I have to say about his “evolutionary informatics” tripe, a tiny complaint. It would be a fine thing if Bob Marks had caught on to the fact that he’s no better qualified to pronounce on evolutionary biology than you are on electrical engineering.
People whose “faith” is such that they feel compelled to rail against the very notion of self-driving cars are pretty badly screwed up. However, I doubt that the money for the center is coming from someone who’s expressing his theological views. It’s more likely coming from someone who will lose money (perhaps in the insurance industry) as self-driving vehicle technology is adopted, or from a libertarian who believes that people are going to lose the freedom to drive their own cars. It’s not too risky a prediction that opposition to self-driving cars will be popping up throughout the conservative ecosystem.
I should add another point to my list of beliefs that the Bradley Center types are very concerned to push:
5. Nonhuman animals are not in any way intelligent.
Anyway, to walk back my assertion: Robert Marks is not ignorant of AI. But many of the other folks there know little more about AI than I do.
As for self-driving cars, it would seem sufficient for these folks to argue that while highly useful, they are not really “intelligent”.
Anyway this is off-topic. Back to Dembski and disproof.
Joe Felsenstein,
What is interesting to me is his claim about the law of information non growth. The only source I could find on this was in Russian 🙁 This states that both determinism and random processes will not produce information. He positively states you need a halting oracle. This moves into a discussion about the touring halting problem which is interesting in itself.
The issue with functional information having selectable functional portions of sequence space biasing along improved fitness seems like determinism and random processes. He may have a valid case against your argument.
What does any of this mean?
Rumraket,
He told us it’s Russian, right?
That leaves me out. 🙁
Joe Felsenstein,
Here is the meat of his argument from Mind Matters. I posted this op in a comment at Uncommon descent.
Touring Halting is bad for the economy.
The scary thing about Touring is that it is not possible to predict whether you will ever be able to halt. Even if you’re riding in a Touring Machine.
Joe Felsenstein,
Nobody does grand touring these days!
Joe, I liked your article at PT. Short, concise, on point. You’re not showing your age at all. 🙂
I am organizing a Turing de France. Anyone interested?
and colewd also quotes Holloway as writing;
So how does any of that work? If we have genotypes, and those have phenotypes, and those have fitnesses, it would seem possible for those higher fitnesses to result an increase of the genotype frequencies of the genotypes whose fitnesses are higher. And that, precisely is higher specified information (SI), And if it gets high enough, it qualifies as CSI.
How are these theorems about mutual information supposed to connect to this? How are they supposed to prevent those changes in genotype frequencies?
(Of course if you redefine CSI as only being there if ordinary evolutionary processes cannot bring it about, then that solves the problem, but at a terrible cost).
Someone needs to connect-the-dots for me as to how these information theorems connect to evolutionary processes. Perhaps Holloway can, but if he does not, one of the other critics of evolutionary theory might give it a try.
Joe Felsenstein,
Eric posted at UD to you and me. I think his hypothesis is that the observation of CSI is evidence of a halting oracle. This is based on the Law of information non growth. If you read his attachments you will see a high level description of how a halting oracle can create CSI and the mathematical proof of how a halting oracle can create CSI.
So how does that work with genotypes and fitnesses? Sorry to be thick about this — perhaps everybody else here sees it clearly.
It doesn’t, but you won’t persuade Holloway of that.
This is classic Holloway. He writes what appears to be an abstract of results that he’s worked out formally, and it turns out that the “abstract” is actually all that he’s got. When pressed for details, as by Mung in the comments at UD (let’s all thank Mung for that), Holloway often reveals that he’s not even clear on the definitions of the mathematical terms he’s slinging. When I say “reveals,” I mean that the revelations of unclarity are to the few people who are clear about the definitions. Holloway has merely acquired vague ideas of formally defined notions, and seems to believe that his manipulation of vague ideas in plain language is a valuable contribution to ID. (I see signs, but can’t be sure, that he regards loose thinking as theorizing, and seeking out evidence in support of one’s ideas as research.) That’s why I characterize the stuff coming from Holloway, and also his BFF Jonathan Bartlett, as mathmaticalistic, and not mathematical.
Sadly, and annoyingly, it’s hard to persuade the folks who ooh and aah at displays of mathematicalism that they’re being duped. To expose mathematicalism for what it is, you have to bring mathematics to bear. But the folks you hope to persuade are not mathematically inclined, and are going to stick with those who tell them the kind of stories that they like to hear.
The reason that this is particularly annoying for me is that, over a period of 12 years, I put a lot of work into cultivation of mathematical rigor in students of computer science. I not only taught 30 sections of discrete math and theory of computation, but also went math-heavy in 33 sections of artificial intelligence. Eric Holloway and Jonathan Bartlett are embodiments of my worst nightmares about educational outcomes. Most computer science students see math as unimportant to whatever work they’ll be doing after graduation, and regard the math components of the curriculum as stumbling blocks thrown in their way. Holloway and Bartlett are much worse than that. They’ve acquired the “look and feel” of the kind of computer scientist that I hoped to cultivate, but not the performance. They talk the talk, but they don’t walk the walk.
Yep. In these cases, it becomes a matter of trust. The duped trust the dupers, and no amount of work you might do will help them out because, well, they don’t even believe that you have something of value to say. They don’t even believe that you can be trusted. Why would they make the effort to try and understand your math?
Thanks. One never knows whether a post like that has made its point clearly, so reactions like that, particularly from the “other side”, are very reassuring.
Tell me about it, as they say. (Fifty years of pushing the stone up the mountain in genetics and genomics, seeing it far too frequently roll back down). As I often remind myself, I volunteered for that assignment so I can’t complain that I was forced to do this.
I have no doubt that Holloway sincerely believes in his argument. I am asking how it actually works, with the expectation that its deficiencies will be apparent once those connections are attempted. But one can never know — maybe we will be shocked and astonished as 100 years of theoretical population genetics collapses before out eyes. Maybe. If you want to bet on that happening I will give you very attractive odds.
Yes, he does.
He has a mathematical model, and his proof is for that model. He does not understand the importance of fitting the model to reality.
I don’t put up comments just for show. What I said at PT, indicating that you’d done a fine job of isolating something specific that could be handled clearly, I really meant.
colewd:
Here is my understanding based on Dr. Holloway’s posts at PS and a few exchanged emails:
1. Dr. Holloway relies on a result proved by Levin in section 1.2 of pdf and re-derived by by Dr. Holloway here.
2. Suppose K(x) is the Kolmogorov complexity of x and I(x:y) is the mutual information between x, y. Levin’s result says that no deterministic or random operation on x can increase that mutual information I(x:y).
3. Dr. Holloway then applies this to evolution
4. But is it correct to represent evolution E(.) by a function that operates only on the original organism? To me, that ignores population genetics. It seems to me that evolution differs from Dr Holloway’s characterization in his E(x) both because it operates on a population of organisms and also because it depends on the status of the organism’s environmental niche.
5. I raised this issue at PS; Dr. Holloway replied to me in a private email that such an operation would invalidate his result. However, he said that changes to the organism and changes to the environment were independent and so considering the environment in the function E(x) would be incorrect. I replied that he seemed to be only considering mutations, not the entire set of evolutionary mechanisms. Dr. Holloway said he would write further on this issue in his posts at the above-linked blog.
6. Dr Swamidass also questioned Dr. Holloway’s understanding of evolution by pointing to an empirical example. He uses evolution of cancer cells within the body to claim that evolutionary processes do introduce new mutual information. Hence Dr Holloway’s E(x) model of evolution must be wrong.
7. What about halting oracles? Dr Holloway does admit that we find mutual information increases in nature. Since he believes this is impossible by evolutionary processes, he concludes that new information is the result of intelligence. He says intelligence can employ halting oracles via a form of hypercomputation, meaning intelligence can exceed the capabilities of TMs. He has posted about this at the end of the article on Levin’s result and also in a subsequent article.
Hey, that’s my line! 🙂
How many times can one say the niche adds bias and be ignored?
BruceS,
What is this “Dr. Holloway” garbage? If you want to go that route, then you had better refer to Joe as Prof. Felsenstein, to Neil as Dr. (is it Professor?) Rickert, and to me as Dr. English. I wish to hell you wouldn’t go that route.
Easily identified as slop from the outset. The equalities and inequalities hold up to an additive constant or up to an additive logarithmic term. That’s not just a “little detail” that everyone understands, and hence need not be mentioned. The notion that leaving out such stuff is making the presentation simpler for readers is preposterous. I’ve only glanced over the math. There’s no telling what else I’ll find when I look closer (shudder!).
The “re-derivation” in “Random Information Non-Growth” is gobbledygook. For instance, the following is meaningless (there is nothing that comes before or after than would make it meaningful):
Also, Holloway is playing a shabby little game of mix-and-match. The terms “conservation” and “non-growth of information” do not come from Levin’s (1984) paper. They come from a translation of the title of an earlier paper that Levin wrote in Russian. Assuming that the translation is accurate, Levin changed his terminology after learning English — and for the better, I would say. He’s now talking about “independence conservation” and “randomness conservation.” Funny how little details like that happen to elude Holloway when they don’t fit with the “information” story he wants to tell.
I am somewhat familiar with this paper of Levin — I can tell you that the major section “I. Algorithmic Information,” including minor sections 1 and 2, is the one of interest here — but I need to look at it again. I won’t be digging in today.
Immediately recognizable as bunk. This is related to something that David Wolpert brought up in his review of Dembski’s No Free Lunch, and that I had brought up in my first paper on NFL theorems: the evolution of a lineage changes the environment, and changes in the environment affect the evolution of the lineage. The NFL theorems do not allow for this. Holloway got the matter totally wrong in his 2011 UD post that Joe links to.
To my knowledge, no one but Jonathan Bartlett and Eric Holloway and fifthmonarchyman (whoever that might be) believes that naturalism entails an algorithmic universe. This is just something that Bartlett made up to suit himself. He’s aware that no naturalist has ever said such a thing. He, humble soul that he is, claims to have clarified philosophers’ muddled thinking for them. The notion that discovery of a non-algorithmic computation equates to discovery of supernatural agency is utter bunk.
There’s much more to say, but I need to leave it for later.
BruceS,
Hi Bruce
That is a great summary. I would add that Dr. Swamadass asked Eric to write a computer program to simulate his hypothesis. Eric did so with mixed results.
I completely agree with that sentiment.
BruceS,
Currently Joshua (Dr Swamadass) and Kirk Durston are discussing this. Joshua has a difficult task in establishing cancer is a gain of FI in my opinion.
I admit I originally wrote this as ‘Eric’ in a draft. But when I got to the bit about Dr. Swamidass, I realized I did not know his first name nor does he ever use it in his forum that I am aware of. So I changed the ‘Eric’ to Dr. Holloway to be consistent.
I am trying to present my best approximation to a balanced view, although it is probably clear where my sympathies lie (and I have made them clear Eric in email). It seemed to me that calling one by his first name and the other by his last or by his honorific would not serve the goal of trying to remain balanced.
Joshua — and many people are calling him that in replies.
As I understand Dr Swamidass, he is defining the functionally equivalent sequences to be those driver genes which have been empirically found to underlie cancer’s fitness. This quote from the cancer thread is important, I believe:
[Start of quotation — I dislike italics in blockquote!]
We need to have a good explanation for how this mutual information, these recurrent mutations, arose. The best explanation is that the cancers have common mutations because they both have the same cancer “function.” Cancer genomes, after all, have the same function. Remembering that genotype often causes phenotype, we rationally conclude that the common mutations are those that are causing the cancer phenotype-function. So, therefore, we infer that recurrent mutations are drivers mutations that cause the cancer function.
[…]
So what can cause high mutual information? It very much depends on the what type of mutual information we are talking about. We can grant a few possibilities.
Intelligence, we can presume is a technical possibility.
Common history / Common ancestry (common starting point before mutation added)
Common mutational distributions and mechanisms (neutral evolution).
Common selective pressures (best explanation for most of cancer FI).
Complex interaction between all of above
[end of quotation]
I see the discussion with Dr Durston (!) as stuck on how to define information increase. Also, Dr Durston seems to be questioning whether the genetic changes in cancer are correctly called new functionality.
(BTW, with regard to Dr Holloway, it seems to me that Dr Swamidass’s new mutual information is between two cancer genomes, whereas Dr Holloway seems to be using a genome and its descendants. However, like many others, I find the mapping of Dr Holloway’s math to the biology of evolution to be incomplete at best.)
It seems to be variations on the same theme. Sequence equating with function which is rare and isolated like tiny islands in a vast sea of non-function.
PS You managed to say Lizzie. Kirk, Eric, Joshua can’t be too hard! 🙂 And it is standard to refer to individuals by surname only in formal writing.
Levin’s math is beyond me. I suspect the ‘proof’ in the blog is not rigorous, but I don’t have the background to justify that impression.
However, I am happy to accept the math as correct for the sake of argument. The issues for me are whether that math has anything to do with biological evolution. I just don’t see how it does, even given Dr. Holloway’s explanation.
Well, it depends what you mean by an algorithmic universe. I don’t think the universe computes. But I do think the many worlds interpretation of QM is the best, and it says the universal wave function is deterministic. (So does Bohmian interpretation, FWIW). So I think the universe can be explained deterministically. (There are issues with determinism in GR and perhaps Newtonian mechanics, but hopefully they go away when QM and GR are reconciled).
(ETA: I’m only referring to explanations at the level of fundamental physics. Stochastic models reflecting our ignorance are needed at likely other levels of science.)
The algorithmic issue only comes up in the work being discussed if you accept that biological evolution cannot create new information. If you do, how does it get there in nature? Iintelligence is claimed to be needed and it works because it can go beyond TM capabilities, according to the post.
The blog post cites a Barlett paper which references work on hypercomputation by Copeland, who has suggested it as a possible way minds could avoid being simulatable by TMs. The cited Bartlett paper itself cites some work which references Penrose’s stuff on Godel showing human minds are not simulatable by TMs (a view which is of course widely disparaged).
Right! I had a brain fart on that one. But would people know him by that name here? In any event, I am happy to change all my references to both from now on.
This is really a core misunderstanding by creationists.
To say that it is functional, it to say that it serves its purpose.
Swamidass correctly understands the purpose of the cancer cell is its biological function (to grow, reproduce, etc). Durston presumably sees this as a misfunction, as a failure to serve the purpose of being non-cancerous.
It is the difference between internal and external. Swamidass says it is functional, because it is performing its internal biological function. Durston says it is not functional, presumably because it is failing to serve the externally ascribed function of being non-cancerous.
And that’s where ID always fails. ID is about external design imposed on a biological system by an external designer. Evolution is about internal design/redesign by the population itself.
Neil Rickert,
Have you read Durston’s last post?
Re: “mutual information”
In the case of evolution, mutual information between what and what?
colewd,
Bill, if you are planning to post at UD on the Eric Holloway thread, you might let gpuccio know that the commenter, Antonin, seems to have been booted from the site by Barry for unexplained infractions.
He was asking here..
Yes, but I’m not sure what to make of it.
We see Durston wanting to impose his own external requirements. In the meantime, the cancer is doing what cancers do.