|
Author
|
Topic: Evolution as a Smolin Class 6 Problem
|
James A. Barham
Member
Member # 50
|
posted 17. June 2002 08:09
I don't know how many of you read Jaron Lanier's wonderful critique of "cyber-romanticism" entitled "One-Half of a Manifesto" in Wired magazine a year and a half ago (Dec., 2000, Vol. 8, No. 12, pp. 158--179), but I heartily recommend it to all who are interested in the problem of defining intelligence and in the problem of the relationship between machines and living things. It was a real breath of fresh air.
I have just discovered a debate that Lanier's article generated on John Brockman's Edge web site, and there was one section by Lee Smolin that seemed extraordinarily pertinent to our concerns here at Brainstorms. In this short piece, Smolin makes some crucial distinctions about design spaces, optimization problems, and classes of algorithms to solve them, that I have not seen made in this way before.
The whole debate may be accessed here:
http://www.edge.org/discourse/jaron_manifesto.html
I also throw in a short exchange between Smolin and Rodney Brooks for its entertainment value.
"From: Lee Smolin
Date: September 27, 2000
Jaron is raising some very important points that deserve closer examination and discussion. Among them is his challenge to the idea that the optimization of present day computers could produce anything with the capabilities of living, intelligent animals, cats let alone people. I think Jaron is right to point out that the arguments for this thesis rest on incorrect assumptions. I believe that Jaron's argument can be strengthened and I would like to explain how. The following is just a sketch, but I hope it suffices to stimulate the debate.
The problems to be addressed are 1) what kinds of problems can computers solve and whether they differ in kind from the kinds of problems humans solve. 2) What kind of problem is it to design a computer and whether it differs in kind from the problem of designing a human, or a creature with equal capabilities.
To approach these questions it helps to begin with the idea that some design problems involve searching a space of possible design parameters. We know that in these cases there are simple optimization algorithms that will find the local extrema in whatever basin of attraction one happens to be in. However, optimization is a small part of design because it can be used reliably to solve only a small subset of possible design problems. To talk about this we may distinguish five classes of design problems.
CLASS 1: Local optimization problems problems which can be solved with standard hill-climbing techniques.
CLASS 2: Locate a pretty good, but not necessarily global extremum in a configuration space with many local extrema and many basins of attraction.
CLASS 3: Locate the global extremum in a configuration space with many local extrema and many basins of attraction.
CLASS 4: Find local extrema in a landscape which changes unpredictably on the same time scale it takes to find local optima.
CLASS 5: find local extrema in cases in which the computation time required to construct the configuration space and/or calculate the fitness function is either infinite or much longer than the time available. These are the class of problems which have to be invented or discovered before they can be solved, as there is no algorithm that can lead to their formulation or complete specification.
Let us first discuss the first question. At least so far, computers are very good at solving CLASS 1 problems, and there are decent algorithms for simple CLASS 2 problems. But we do not have good methods for finding global extrema and hence solving CLASS 3 problems. To my knowledge computers can do decently at some simple CLASS 4 problems, but can easily fail when they become more complex. By definition computers have problems solving CLASS 5 problems, as the computation time to set up the extremization problem is prohibitive. However humans can often solve CLASS 3 problems and are also quite good at CLASS 4 problems. This should be no surprise, this is part of our biological specialization. This is what is required to flourish in a new environment, domesticate a new species, become farmers, populate almost all the ecological zones on the planet and so forth.
But humans can do even better than that, we can both invent and solve CLASS 5 problems. This is what poetry, art, music and science, are about. We invent the forms and traditions and then we master them. We can thrive in a domain in which we create optimal versions of things that did not even exist a short time before. We are not extremizing in a landscape, we are building the landscape on the same time scale that we master it.
One correspondent suggested that anyone who thinks people are different from machines are naive romantics. This is not true, we are different because we have vastly different capabilities. It is irrelevant to talk of the universality of Turing machines, for Turing machines are entities that run programs that must be written by an external entity. So far at least the only entities we know of who can function as those external programers are humans. Humans are intelligent creatures that do not need to be programmed by any external agency. Turing machines are designed, we are the result of natural selection. We need then to examine the second question, whether designing or programming a computer is in the same CLASS of problems as the problems natural selection solved in the course of evolution.
Of course inventing the idea of a digital computer was a CLASS 5 problem. But once we had the idea, the optimization of digital computers is mainly a CLASS 1 problems. This is what Moore's law is about, it tells us how quickly local optimization can work when ample resources are available. One of the points Jaron is making is that the design of software required to do justice to the exponentially increasing capabilities of our machines are not CLASS 1 problems. Moore's law tells us that the fitness landscape for software is changing on a time scale comparable to the time required to write and debug software. Thus writing software involves problems of at least CLASS 4. This is of course just a different way of making one of Jaron's arguments.
For there to be a danger of robots taking over, or even being able to do a decent job entertaining us, replacing songwriters and singers,artists, scientists and comedians, one of two things have to happen. Either we will be able to design a machine that could replace us, which means a machine that can solve problems of CLASS 5, or we will be able to design a machine that could in turn design a machine that could solve CLASS 5 problems.
But while we can solve problems up to CLASS 5, so far we have only been able to design machines that can solve CLASS 2 problems reliably. And so far machines are not able to design other machines to solve even CLASS 1 problems. When one puts it this way it is clear that it is not just a matter of Moore's law, designing one of us is a very different kind of problem then optimizing a programmable digital computer.
What kind of problem is it to design an entity that can solve CLASS 5 problems? We know we were created by natural selection, acting on not only us but the whole collection of living species. This is at least a CLASS 4 problem, but it is very likely at least a CLASS 5 problem. The interactions among many species as they evolve under the rules of natural selection is a CLASS 4 problem, as is shown by models of Bak and Sneppen, Kauffman, Sola and others. But there are good arguments, summarized in Stuart Kauffman's forthcoming book, that natural selection and cultural evolution are really CLASS 5 problems. He argues that they are problems in which the construction of the fitness landscape itself is so computationally intensive that it is not correct to separate the specification of the fitness landscape from its optimization. Instead, both take place together. This means really that the metaphor of optimization has broken down completely. Whatever evolution is doing cannot, he argues, be conceptualized as extremization on a pre-existing fitness landscape.
Thus, the problem of designing an entity that can solve CLASS 5 problems is at least a CLASS 4 problem, and very likely is a CLASS 5 problem. But is it only this hard, or harder still? Human's can solve some CLASS 4 and 5 problems, but it is not at all obvious that the problems of these kinds that we can solve are comparable to the problems that natural selection has solved in designing us. At the very least, it is likely that the time required to solve the problem of designing us may take a great deal longer than the tine it takes to solve the CLASS 4 and 5 problems we have so far dealt with. It took natural selection 4 billion years to design us. Let us assume that we could do it much faster. How much faster? Let us assume that we could use genetic engineering to engineer an artificial speciation in an animal. Speciation is a process that takes on the order of 100,000 years. Given very optimistic assumptions it is possible to imagine that some years from now this is something we will be able to accomplish in on the order of 100 years. It could certainly not be less than that as we cannot do it faster than the time it takes for several generations to grow to maturity. (Because the interaction of an animal and its environment is a CLASS 5 problem, we are not likely to be able to simulate it reliably enough to replace the phase where we grow the animal and observe what happens.) This would mean that we had the tools to speed up natural selection by a factor of 1,000. Even with this fantastic increase of speed it would still take us a million years to invent something like ourselves, starting from scratch. (Note that this is true even if we skip the pre cambrian stages of evolution, which begins with creatures whose cell biology and biochemistry is far advanced of what we have so far designed. Note also that many biologists working in parallel won't help as natural selection also works in parallel.)
This is on the order of the lifetime of a species. A problem like this, whose minimum time for solution is on the order of the lifetime of a whole species of creatures that can solve CLASS 5 problems deserves a separate class. So we may call this a CLASS 6 problem.
Is it possible that there is a way to do it much faster, by taking a route that natural selection could not have? One cannot say this is impossible, but all this means is that so little is known about the problem that it is in a class of problems we have no idea how to solve.
To summarize: the claim that optimization of present computer designs could produce something that is "as powerful" as humans requires that there is only one kind of intelligent entity, and they all live in a in a fixed landscape with a single local extremum. But we are not only not in the same basin of attraction as present day computers, it is not even obvious that the problem of constructing us has anything in common with problems we have so far solved. This is not to deny that someday humans may learn how to solve the problem of designing creatures that can themselves solve CLASS 5 problems. The point is only that there is no rational basis for predicting when or even whether this may happen, as the solution to this problem is not closely related to the kind of optimization problems that human designers have so far learned to solve.
From: Rodney Brooks
Date: October 1, 2000
Lee Smolin wrote:
"One correspondent suggested that anyone who thinks people are different from machines are naive romantics. This is not true, we are different because we have vastly different capabilities. It is irrelevant to talk of the universality of Turing machines, for Turing machines are entities that run programs that must be written by an external entity."
This is exactly the sort of naive romanticism to which I was refering. I was not comparing humans to a PC running Windows 2000. I am saying that people are machines in the sense that there is, as far as we have any scientific knowledge at this time, nothing in them outside the laws of physics of the universe which govern all matter. People are made of matter and that matter obeys the physical laws of the universe. Unless one hypothesizes an eternal soul, an elixir of life, an ineffable essence, or some other extra-physicalness to humans (and also to other animals, all the way down to bacteria?), then humans are machines. It has absolutely nothing to do with Turing machines, or programming computers.
Get over your fear of being a machine. We are not the center of the universe, and God does not exist. That is what this disagreement boils down to.
From: Lee Smolin
Date: October 2, 2000
In reply to Rodney Brooks:
I believe strongly that our entire existence is as part of the natural world. I am not afraid of this; my book, The Life of the Cosmos, is a kind of homage to that idea. My guess is that we agree broadly on metaphysics, but my comment had nothing to do with God, cosmology, consciousness or any kind of romanticism. I was trying to make a point about science, one that is well within the boundaries of our shared metaphysics.
In my comment I raised two issues. First, whether everything that is part of the physical universe can be described in terms of a Turing machine, second, whether the way that living animals process information is enough like how digital computers work that it is rational to hope to construct a reasoning animal based on models of digital computers. As these seem to be very open issues given the present evidence, it seems far from clear that the metaphor of a machine will in the end be very helpful to us as in understanding in physical terms what animals are. In addition there is a problem with using the word machine in this context, which is that it carries with it the implication that something was made by human beings. This is not just semantics because ignoring the deep differences-as physical systems-between living animals and human made machines has led to some predictions for the future of machines that may not be consistent with our developing understanding of what life is.
To expand on this last point, I do believe that we will someday understand what we are in terms of physics. But before we do that we must first understand what a living thing is in terms of the laws of physics. We have made a lot of progress towards this in the last years and I believe more will be made shortly. Everything we have learned suggests that there are important differences, expressible in completely physical terms-more particularly in terms of statistical physics, between systems that are made and systems systems that arise by a spontaneous process of self-organization. Both may process information, but they may do so in different ways, so that they are generally able to solve different classes of problems.
A related point is made by Stuart Kauffman in recent papers and a forthcoming book: there is a fundamental difference between a physical system that can be termed an "autonomous agent" and one that cannot be. Part of Kauffman's definition of an autonomous agent is that it is a self-reproducing system, able to carry out at least one thermodynamic work cycle. Computers are not autonomous agents to the extent that they are constructed and programmed. But computers are Turing machines-which is why that idea is useful for this discussion.
Living animals are autonomous agents. They are not, so far as has been shown, Turning machines. There is no obvious relationship between the definition of a Turing machine and the definition of an autonomous agent; it is certainly very unlikely that they are equivalent. Thus, while it is of course possible that we may some day be able ourselves to make living things, there does not seem to be any good reason to expect that such articial animals will have a strong resemblance structurally or functionally to computers. (The fact that one can model certain aspects of life in computer software does not change this.)
Computers are wonderful tools and fantastic toys. But if machine is to mean anything at all besides "something found in the universe" (remember that we have the same metaphysics) then computers are machines, and animals are not."
JB:
I also include a couple of wonderful quotations by Philip W. Anderson from the same debate, which summarize what I feel to be the chief issue with just the right kind of the quiet good sense:
"From: Philip W. Anderson
Date: October 16, 2000
I was very happy to see Jaron Lanier's paper, in that it was saying a lot of things I had felt to be true, and saying them from within the digital world. . . .
I ran across a quote from, oddly, G. K. Chesterton, which makes one of the points nicely. 'life is a trap for logicians; it looks just a little more mathematical and regular than it is. Its exactitude is obvious, but its inexactitude is hidden; its wildness lies in wait.' . . .
I guess the problem I have is that discrete mathematics feels too anthropomorphic - too much creating the world in our own image. No matter how far Moore's law carries us, it is still digital. I am not agreeing with Penrose, nor do I believe we are anything but a machine - but are we a digital machine? To put it less mystically, is a digital representation practical?"
JB:
It seems to me that the foundational problem with respect to intelligence is the distinction between organisms and machines. Posing the problem of the essential nature of life in terms of whether only organisms can solve Class 6 problems, or whether machines might some day do so, was a particularly fruitful thing to do, I felt.
While I am sure that neither Smolin nor Anderson would be willing to accept the label "vitalist," nevertheless that is the crucial issue, and the debate seems to be slowly heading in that direction.
[ 17 June 2002, 23:30: Message edited by: James A. Barham ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 17. June 2002 10:11
James,
I couldn’t access the web-site so my comments are based on the materials you posted. Let me start by saying that, 1)the discussion addresses what is clearly one of the central issues in the area of natural versus machine intelligence, and 2)(IMO) the way they have formulated the problem essentially guarantees they will never find a solution.
THE PROBLEM
FACT 1: As far as I am aware, almost all artificial intelligence is based on the idea of ‘finding or searching for optimal solutions’ to problems. Whether based on a Darwinian genetic algorithm or some other search routine, the more complex the problem, the more processing time(the more lifetimes) required to solve the problem.
FACT 2: Human and animal problem solving appears to be based on very different concept or approach. Although humans have a great deal of difficulty in solving some types of problems, we can solve certain extremely complex problems very, very quickly. In fact, humans solve many complex problems in a single trial.
Given the these two reasonably well documented facts, the AI community has concluded that the key to simulating intelligence is a more powerful search algorithm. Given the same facts, the ID community has concluded that machine simulation of human intelligence is logically impossible. The adaptive approach concludes that everybody else is looking at the issue from the wrong perspective. [May the moderators forgive a feeble attempt at humor.]
The adaptive approach suggests that the reason human, animal, and cellular intelligence look so different from machine or artificial intelligence, is that 1)they are not attempting to solve complex optimization problems and 2)they are not attempting to find optimal solutions. The adaptive approach suggests that human, animal and cellular intelligence are used to solve relatively simple problems by systematically searching a relatively limited range of options in order to find a solution that is adaptive rather than optimal. I have developed a rather interesting little experiment to demonstrate that this is, in fact, how human intelligence operates.
Intelligence as ‘the ability to very rapidly find adaptive/non-optimal solutions to a wide range of adaptive problems’, produces, almost as an incidental side benefit, the ability to generate novel and creative solutions to problems. The tendency or ability of an ‘intelligent system’ to optimize both reduces the ‘intelligence’ of the system and its ability to generate creative solutions.
I am well aware that the above conclusions directly contradict almost universally accepted conventional wisdom. However, I stand ready to demonstrate these conclusions both mathematically and experimentally.
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 17. June 2002 14:59
quote:
James Barham: It seems to me that the foundational problem with respect to intelligence is the distinction between organisms and machines. Posing the problem of the essential nature of life in terms of whether only organisms can solve Class 6 problems, or whether machines might some day do so, was a particularly fruitful thing to do, I felt.
While I am sure that neither Smolin nor Anderson would be willing to accept the label "vitalist," nevertheless that is the crucial issue, and the debate seems to be slowly heading in that direction.
I think you have misunderstood the intent behind Smolin's remarks. It seems to me that Smolin was not criticizing the ideas that human intelligence is (i) computable, (ii) a product of physics, and (iii) in princinple understandable in terms of already existing physics models. (Perhaps he doesn't accept all of these ideas, but I don't think the purpose of this particular exchange was to express his disbeliefs, if any, about this.) He was merely trying narrow down the meaning of term "machine". My interpretation is that he thinks that the term "machine" should be confined to the kind of objects that are similar to our past and present technology, and that living organisms are not similar to our technology. To be a vitalist he would have to adhere to some mysticism regarding what makes living organisms alive, which is something differently than merely trying to tidy up the semantic swamp.
If I'm right, then I agree with Smolin. There really is a disanalogy between living organisms and machines. A living organism is as much a process as it is a thing, while our technological objects are just things. But I can take this position and still believe that, e.g., human intelligence is ultimately reducible to some Turing machine without the risk of inconsistency.
I am bit more uncertain about Smolin's classification of problems. I'm sure there is a heuristic value in such a classification, but from a more theoretical point of view it seems a bit arbitrary (for intelligence; it looks good for evolutionary optimization). I think intelligence is best thought of in terms of learning rather than optimization. We humans are faced with a lot of sensory data every day, and our task is to make the appropriate decisions based on this data. For this we must learn and predict. While the process may also involve optimization, I think learning is the central thing. Therefore, and with Occam's Razor Theorem in memory, I think WinZip is a more interesting model for intelligence than the hypothetical computerized class 5 problem-solvers.
Erik
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 17. June 2002 22:53
Erik:
Of course, as I said, I realize that Smolin would never himself accept the label "vitalist". But what does it mean to claim (as he does) that machines and organisms belong to fundamentally different classes, with fundamentally different capabilities?
It means that there must be a sui generis physics of the living state (presumably deriving from the huge size of macromolecules giving rise to frustration and metastability, as well as some sort of long-range coherence of the protein-water gel).
But whatever the explanation ultimately turns out to be, the crucial question is whether the material constitution of life is what endows life with its special properties, or whether life and mind are simply a matter of getting a certain abstract organization right. That is to say, is the doctrine of "multiple realizability" correct, or isn't it? It seems to me that Smolin is pretty clearly denying it (but I guess I could be wrong).
If I am interpreting him correctly, then I agree wholeheartedly. What we choose to call this belief is not that important, but the CRITICS of the belief will certainly call it "vitalism," because the word is so pejorative. But they would be justified in doing so, because one of the meanings of that term is the claim that there is something special about life.
Now, as for me, I call my position "biofunctional realism"---meaning that the teleology or intelligent agency manifest in life is an objectively real, emergent feature of the world. I believe, as I said, that it is a direct result of the special physics of the living state that we do not currently understand. Whether Smolin would go quite that far, I doubt.
As for the word "vitalist," once I got up the courage to admit to myself that it is an accurate description of my position, I finally began to embrace it. There's no use pretending to be something we're not. I am a realist about biological functions. That means I believe life is special, in ways we do not yet understand. Therefore, objectively speaking, I am a vitalist, under one long-standing and well-accepted meaning of the word.
Finally, I think that Smolin's and Anderson's positions are similar to mine, even if they would not wish to go as far as I do, or use that word. Thus, I am very encouraged.
[ 17 June 2002, 23:25: Message edited by: James A. Barham ]
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 17. June 2002 23:07
Warren:
Sorry about the problem with the link. I think it's working now (and thanks to Iain for telling me what the problem was).
I am not quite sure what to say about the relationship between Smolin's classification of optimization problems and your own ideas.
As I understand it, I think you are right about the local nature of most goal-directed problem solving. That is, I think you are right that what living things mostly do is satisfice, not optimize in the strict sense. On the other hand, there is something about the intelligent agency of life that does lead to extremely good solutions that are in some cases nearly globally optimal (the retina is an oft-cited example). So, I don't know that I want to exclude the global optima categories from consideration altogether.
As for the relationship between your algorithms and Smolin's landscapes---at least with respect to the local optima---is there any necessary conflict there? Could they be two ways of saying the same thing? I have the impression that these are formalisms that are largely intertranslatable? (But correct me if I am wrong.)
[ 17 June 2002, 23:33: Message edited by: James A. Barham ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 18. June 2002 10:30
James,
QUOTE: As for the relationship between your algorithms and Smolin's landscapes---at least with respect to the local optima---is there any necessary conflict there?
My observation is that there are significant differences between 1)neo-Darwinian systems designed to find local optimums(LO systems), and 2)systems designed to find multiple adaptive solutions to multiple problems in a highly dynamic landscape(MA systems).
The first, the two types of systems involve different functionality. You can design LO systems using little more than the NDG mutate-select functionality. MA systems need additional functionality to perform actions like retrieve/access solutions and preserve diversity. It is not possible to model or simulate biological systems without introducing functionality in addition to the mutate-select functionality needed to find a single optimum.
Second, major difference is behavior. Like real biological systems, MA systems don’t optimize even simple problems, and they find solutions very, very quickly. These are both behaviors that clearly differentiate (biological and MA systems) on the one hand and (neo-Darwinian and LO systems ) on the other hand. Understanding the ‘reason’ for these behavioral differences is, IMO, key to understanding the fundamental flaws in both Darwinian evolution and traditional AI as described by Smolin’s landscapes.
One reason for the behavioral differences is processing capacity. It takes a lot of energy to find optimal solutions, and in most instances, an optimal solution has very little advantage over a non-optimal adaptive solution. If a system or organism must solve many billions of adaptive problems per lifetime, it is important to make efficient use of processing capacity.
A second reason for non-optimization is diversity maintenance. If the solution to an adaptive problem is likely to be different tomorrow or in 5 milli-seconds, it is useful maximize diversity to increase the likelihood of quickly find an adaptive solution to a new set of conditions.[One of the key ‘absurdities’ of NDG is the claim there are no processes in biological systems to actively maintain and maximize ‘diversity’]
TELEOLOGY
It is interesting to note that LO and MA systems share one important feature. Both are systems to find purposeful, teleological, or adaptive solutions or causal relationships. The neo-Darwinian suggestion that teleological causation is not desirable in science, is, IMO, largely an argument that the neo-Darwinian model is the only acceptable teleological explanation.
QUOTE: On the other hand, there is something about the intelligent agency of life that does lead to extremely good solutions that are in some cases nearly globally optimal (the retina is an oft-cited example).
The differences between LO and MA systems as discussed above involve differences in how the two types of systems find elementary solutions to elementary problems. As you point out, ultimately, the differences between the two systems must be addressed in terms of how well they explain ‘observed instances of biological systems finding solutions to very complex problems’.
There are, as you point out, clearly examples where nature appears to have found near optimal, and highly innovative solutions to extremely complex problems. It is useful to ask whether LO systems or MA systems can do a better job of explaining such complex problem solving.
In favor of LO systems, there is ample evidence that 1)given the proper set of conditions, 2)given enough time, and 3)given a systems set up to solve the specific problem, that complex forms of LO systems can generate the types of complex near optimal solutions found in nature. The obvious weakness in the LO/Darwinian explanation for the creation of complex solutions, is explaining how all the required prerequisites could possibly come into existence for all the instances where nature has generated complex near-optimal solutions to problems.
It is easy to demonstrate that under appropriate conditions, MA systems like LO systems will find near optimal solutions to complex problems although techniques involved are somewhat different. MA systems use parallel processing. In a MA system, a complex problem is represented as a set, typically a hierarchical set of simpler adaptive problems. Solving the simpler problems creates the appearance of having solved a single complex problem.
Problem in MA systems, compared to problem solving in LO systems, is believed to be much more like the problem solving performed by biological systems. It can, I believe, be demonstrated that 1)MA systems solve complex problems much faster than LA systems, 2)they are capable of far more ‘creative’ problem solving, and 3)MA systems can explain the set up or initialization of complex problem solving(In theory, certain types of MA systems can change or evolve to address any complex adaptive problem.)
MA systems are complex, and evaluating the validity of the claims made for these systems will be no simple task. My objective in introducing these concepts into this discussion are to point out that
1. Simulating/modeling human problem solving and simulating/modeling biological evolution are very similar problems involving many of the same issues.
2. Conventional wisdom in both evolutionary biology and AI suggest that the best, and in fact the only way, to develop a model or simulation is to start by addressing the question "how does the organism or system find the optimal solution to a single adaptive problem". I am suggesting that this is neither the only nor the best starting point. If you want to understand biological problem solving, you need to start with the question "How does an organism or system quickly find adaptive solutions to a whole array of different adaptive(teleological) problems?". And
3. The current ‘find optimal solutions’ models are dead ends for both evolutionary biology and AI. Simple ‘find optimal solutions’ models don’t work. The is true of both simple RM&NS models and simple GA models. By making these models more complex, it is possible to solve more complex problems, but these problem solving systems diverge rather dramatically and quickly from what is observed in biological systems.
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 18. June 2002 17:54
Warren:
Thanks for your further efforts to clarify the distinctions between your "mutiple adaptive" algorithms and standard hill-climbing algorithms. I wish I had the technical expertise to follow everything in detail, and probe more deeply, but alas I don't.
Best of all would be if Smolin himself (or someone else who understands his claims well) were to enter the discussion! We need someone capable of comparing and contrasting the two approaches more effectively than I am able to do, to give you a worthy interlocutor . . .
IP: Logged
|
|
|
Printer-friendly view of this topic