|
Author
|
Topic: The Characterization of Intelligent Causation
|
LifeEngineer
Member
Member # 3446
|
posted 02. April 2007 08:10
Quote:You seem to be saying that we should characterize "intelligence" as "that which is Turing-computable". In other words, in answer to the question "How does a human being design a watch?", the answer sholud be "By means of some process that is Turing computable". Given that everything is Turing computable as far as we know, you are saying that everything is intelligent.
That clearly is not what I said. First, I have defined a logical or mathematical structure or framework for defining information, intelligence, and so forth. I have not yet proposed a defintion of intelligent causation, but if you understand the framework being defined, it should not be too difficult to anticipate the definition.
Second, the abstract logical structure or framework presented is not simply a Turing machine. In addition to the Turing type information processing machine, the mathematical framework defined includes 'search areas' and solution spaces.
I have presented a fairly standard abstract mathematical framework or structure involving 1)input variables, 2)processing algorithms, 3) processing, 4) output variables, and 5) search areas and solution spaces for modeling the success or failure of the 'problem solving' modeled and simulated by a Turing machine.
As anyone familar with abstract modeling should recognize, I am presenting a precise mathematical/logical structure or framework within which a more formal definition of intelligence as problem solving can be formulated.
For those with a working knowledge of abstract modeling, it might be useful to note that I am presenting a fairly standard abstract framework that including the abstract components of an abstract logic machine (Turing machine) and an abstract structure outside the machine that makes it possible to model the success or failure of a Turing machine at solving a problem or achieving a goal.
As should be apparent or obvious, the framework defined can be used for modeling any electronic computer operation or program. The framework can thus be used to model any behaviors modeled or simulated by computer programs. This includes behaviors like human behaviors and biological evolution behaviors. It also includes things like the behaviors of hurricanes.
To address aiguy's point further, the claim will be made that the framework defined can be used to define and model intelligent behavior. But, as aiguy points out, the framework framework would be meaningless unless it could also be used to define and model non-intelligent behavior.
It is useful to note that abstract mathematical/logical models represent or express only a very limited subset of real world properties associated with the phenomenon being modeled. It is important to recognize that the framework or structure defined here does NOT model or express phenomenon like Gods or designers or programmers or minds or consciousness. A key part of the difficulty a lot of people have in understanding abstract scientific modeling, is understanding how and why scientific modeling can exclude so many potentially relevant variables.
Finally, it should be noted that 1) you can not develop meaningful scientific definitions without a clearly defined framework and 2) the framework defined here can be described or characterized as a common or standard or widely used framework for analysis. It might, in fact, be characterized as a framework that is applicable to all computer programs and all analysis performed using computers and computer models.
IP: Logged
|
|
LifeEngineer
Member
Member # 3446
|
posted 02. April 2007 08:53
Having defined a logical/mathematical framework, and one that is apparently widely used and accepted, the next step, before offering a definition of intelligent causation is to outline the criteria to be used to evaluate and accept or reject the definition proposed.
We can start with two of the criteria suggested by aiguys comments.
First, the proposed definition of intelligent causation must be applicable to both human behavior and biological evolution. As a side benefit, the definition to be offered will also be applicable to intelligent computer behaviors.
Second, the definition must define or diferentiate both intelligent and non-intelligent behaviors. The definition to be offered will define and permit (relative) quantification of degrees of intelligence.
Third, the definition must result in or be associated with non-trivial testable predictive theories. In fact, the definition must also be compatible with or associated with complex sets of logically consistent predictive theories.
Fourth, the definition must be judged or evaluated in relationship to competing definitions that also produce or are associated with non-trivial predictive theories. A scientific definition does not have to be perfect. It simply has to be current best practices associated with current best practices predictive theories. It should be noted that no alternative or competing formal frameworks for defining intelligent causation are being presented or discussed here.
Finally, the definition must be judged acceptable by those actually producing and testing non-trivial predictive theories.
Science or hard science can be characterized as a complex form of human behavior with a demonstrated capability to produce the most efficient best practices problem solving. Science can also be viewed as a method of determining the current best practices definitions, predictive theories, and practical solutions to problems. In order to apply or practice hard science evaluation, it is often important to first define the logical framework or structure being used and the formal scientific standards and criteria involved in practicing scientific evaluation.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 02. April 2007 10:15
Melvin: quote:
Look, call me when you tell the computer to guess because you have programmed it to guess and its response is to print to screen; “No! Guessing is wrong.”
How about if you call us when you find anything that can be shown to exercise the libertarian free will that you're talking about.
IP: Logged
|
|
aiguy
Member
Member # 3736
|
posted 02. April 2007 12:28
Mel,
quote:
These systems gather information as they have been told to do and store it where and how they have been told to do.
Pretty much like children are told to store knowledge, really. Does that mean that humans are programmed?
quote:
Novel solutions, sure, but the search algorithms are prescribed and any modifications made by the computer to these algorithms are done in a prescribed way.
Again, the way our learning algorithms are determined. But we learn, and come up with novel solutions. How do you think this is qualitatively different from what we do? Or, more to the point, how do you think we could demonstrate that this is qualitatively different from what we do?
quote:
Guess? This is an interesting choice of word. The computer has no inkling of what a guess is.
Yes, they do. We usually call it "heuristic" programming, but when we are talking about what some system has done, we talk about just the way we would if discussing a human's performance. "Why do you think it decided to do X? Probably because it wanted to get some information on Y, and didn't know how to do Z, so it was guessing that X was the best way to go in this situation..."
quote:
It is told exactly how to select what to do. This selection may be as a result of some “random” process as set down in the programming but the computer can’t differentiate those instructions from the type of instructions we would classify as a non-guess.
Yes, they do. We can't possibly tell complex AI systems what to do in any situation, because we have no idea what situations will arise. Computers try different things, just like we would, in a situation where there is uncertainty involved.
quote:
Look, call me when you tell the computer to guess because you have programmed it to guess and its response is to print to screen; “No! Guessing is wrong.”
Watch, I'll make your computer say that: GUESSING IS WRONG! ![[Wink]](wink.gif)
IP: Logged
|
|
aiguy
Member
Member # 3736
|
posted 02. April 2007 12:30
LE,
If you have some actual proposal for a characterization of intelligence that would enable us to test the identity relationship in question, let's hear it. So far you've said only that it can be modelled on computers, which doesn't help.
[ 02. April 2007, 12:32: Message edited by: aiguy ]
IP: Logged
|
|
LifeEngineer
Member
Member # 3446
|
posted 02. April 2007 13:13
Aiguy,
Does this mean you are in agreement with the framework and criteria defined? We might also ask if you actually support some alternative framework, criteria or characterization? It is generally easier and more productive to evaluate the relative merits of competing definitions and approaches than it is attempt to defend one definition against a lot of meaningless arbitrary criteria that can not be satisfied by any definition.
IP: Logged
|
|
aiguy
Member
Member # 3736
|
posted 02. April 2007 14:14
LE,
quote:
Does this mean you are in agreement with the framework and criteria defined?
Let me try this just once more. The question at hand is this: How can we characterize "intelligence" such that we can show that the cause of human behavior and the cause of biological complexity is actually the same thing. This characterization must be at some suitable level of abstraction. For example, if we simply said that intelligence was "that which creates complex designs" or that intelligence was "problem solving", we obviously could not demonstrate the truth of the identity relation that ID proposes, since that would be too abstract. On the other hand, if we characterize intelligence in terms of the activity of human neural processes, that would presumably be too concrete to allow a comparison to the cause of biological complexity, since presumably the cause of biological complexity does not actually depend upon a human brain.
Your suggestion, if I understand it, is to characterize intelligence as "that which is Turing computable". Again, since as far as we know everything is Turing computable, there is nothing that theoretically can't be modelled by a Universal Turing Machine (UTM). We can model turbulence in air flow, or the path of a comet, or raindrops falling on your head, all with a UTM, but nobody thinks that these are all examples of the same thing.
So until you provide some model that is more specific than "Turing-computable", we have nothing to go on.
quote:
We might also ask if you actually support some alternative framework, criteria or characterization?
I have already answered this: I am aware of no unitary characterization of human intelligence at all. In AI, there are a large number of techniques used to replicate specific mental abilities, and for some mental abilities (e.g. understanding unrestricted natural language at human-level competence) we have no successful techniques at all.
[ 02. April 2007, 14:26: Message edited by: aiguy ]
IP: Logged
|
|
Melvin H. Fox
Member
Member # 1684
|
posted 02. April 2007 17:41
Calling 2ndclass and aiguy,
It is clear to me that neither 2ndclass nor aiguy have children, at least none like mine. I have tried to program them and indeed some of the code sticks and works very well, but sadly, some of it is repeatedly rejected. Rejected by what, the operating system? Fine, for now we will call it the operating system instead of that horrible devil – free will – which seems to puzzle you all. Where did this operating system come from and how does it work?
You would say it is a heuristic system (negative circular processing) and not algorithmic (solution certain) and perhaps some of it is. After all, the scientific method has proved very valuable in human advancement. But if you are correct and humans operate solely based on this glorified trial and error system then where did the original blue print for guess and check come from and who or what defines the acceptable limits for a perspective solution or answer? How is right enough and too wrong established?
Your question: “How can we characterize "intelligence" such that we can show that the cause of human behavior and the cause of biological complexity is actually the same thing?” Since it is doubtful we will agree any time soon about the workings of a computer and the workings of the human mind I will redirect my efforts toward the above query.
I caution you not to ignore the obvious inconsistencies in human behavior. If you refuse to acknowledge the scientifically undemonstrated free-will as the cause for these inconsistencies, then demonstrate some other cause. One thing we demand of the scientific method is that results must be repeatable. If human behavior is the result of heuristic programming then the defined limits of right and wrong among humans should be much more consistent and should be converging instead of diverging as is observed.
-Mel
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 02. April 2007 18:45
quote:
It is clear to me that neither 2ndclass nor aiguy have children, at least none like mine. I have tried to program them and indeed some of the code sticks and works very well, but sadly, some of it is repeatedly rejected.
Unfortunately, my kids are likewise resistant to programming, which is probably why I find computers to be such a welcome diversion.
quote:
Where did this operating system come from and how does it work?
That's a perfectly good scientific question, and it's being addressed by cognitive scientists and others. Aiguy would be the right person to comment on that, but obviously there's no short or complete answer.
quote:
Since it is doubtful we will agree any time soon about the workings of a computer and the workings of the human mind I will redirect my efforts toward the above query.
Seems to me that the workings of a computer are completely uncontroversial. Any disagreement should be empirically resolvable.
quote:
I caution you not to ignore the obvious inconsistencies in human behavior. If you refuse to acknowledge the scientifically undemonstrated free-will as the cause for these inconsistencies, then demonstrate some other cause.
If lack of scientific evidence were the only problem with libertarian free will, then maybe I could deal with that. My personal problem is that I don't know how to define it in a logically coherent way. I don't see how the categories deterministic and indeterministic leave room for a third alternative.
Inconsistency is hardly unexpected in systems that are ultra-complex and ultra-open, as humans are. I don't know why human inconsistency would require a special explanation any more than we need a special explanation for the fact that the weather often behaves unexpectedly.
quote:
One thing we demand of the scientific method is that results must be repeatable. If human behavior is the result of heuristic programming then the defined limits of right and wrong among humans should be much more consistent and should be converging instead of diverging as is observed.
The problem is that the "programming" is intractably complex, and it's subject to a constant influx of environmental data. I suppose that some aspects of human behavior lend themselves to simple models, but I would expect them to be the exception. And I don't see why a non-free-will hypothesis would predict convergence of behavior.
IP: Logged
|
|
Melvin H. Fox
Member
Member # 1684
|
posted 02. April 2007 20:07
quote:
If lack of scientific evidence were the only problem with libertarian free will, then maybe I could deal with that. My personal problem is that I don't know how to define it in a logically coherent way.
My point exactly. Humans (intelligent agents, and so incomparable to weather patterns), to a large degree, do not behave in a logically coherent way. Therefore, the cause of their behavior may not be explainable in a logically coherent way.
-Mel
IP: Logged
|
|
aiguy
Member
Member # 3736
|
posted 02. April 2007 22:28
Hi Mel,
quote:
It is clear to me that neither 2ndclass nor aiguy have children, at least none like mine. I have tried to program them and indeed some of the code sticks and works very well, but sadly, some of it is repeatedly rejected. Rejected by what, the operating system? Fine, for now we will call it the operating system instead of that horrible devil – free will – which seems to puzzle you all. Where did this operating system come from and how does it work?
I can't even make my dogs behave. But neither can I make my computer systems behave. I try to program them, and sometimes they do what I want, but sometimes they don't, and I do not know why. Don't mistake this for some sort of programming "bug"; the code is in fact exactly as I intended to write it. But in certain situations, the system decides to do things that I would rather they didn't. I think that one rule I've given the system should be what determines its behavior, but the system goes by another rule instead in certain circumstances. Just like kids: There are lots of things that determine their behavior. They are born with some of them, and they learn some too. They don't always follow the rules you want them too, but that's not evidence of some magic power they have to do something besides what their rules, in interaction with their circumstances, tell them to do.
quote:
But if you are correct and humans operate solely based on this glorified trial and error system then where did the original blue print for guess and check come from and who or what defines the acceptable limits for a perspective solution or answer? How is right enough and too wrong established?
I don't know where it came from. Maybe it came from another glorified trial and error system. And where did that one come from? Yet another one perhaps. We can speculate that the very first trial and error system simply popped into existence all by itself, which is a very hard thing to imagine. But even harder to imagine is that something else, which was not a trial and error system, and of which we have no understanding at all, simply popped into existence all by itself, and caused the first trial and error system to exist. Either way, it is clearly beyond our ability to decide empirically the nature of this ultimate cause.
quote:
I caution you not to ignore the obvious inconsistencies in human behavior. If you refuse to acknowledge the scientifically undemonstrated free-will as the cause for these inconsistencies, then demonstrate some other cause.
We understand things scientifically in terms of two types of causes: fixed law, and chance. So what explains inconsistencies in human behavior? Two things: First, even if we were perfectly deterministic machines, a human being need never respond the same way twice, since the totality of our previous experiences plus our innate characteristics would determine our responses to our circumstance. And second, perhaps just chance.
You should note that the same applies to computer systems, which do not necessarily act consistently either, for exactly the same two reasons.
quote:
If human behavior is the result of heuristic programming then the defined limits of right and wrong among humans should be much more consistent and should be converging instead of diverging as is observed.
Sorry, I don't understand this. What is diverging? Why would we predict convergence?
quote:
My point exactly. Humans (intelligent agents, and so incomparable to weather patterns), to a large degree, do not behave in a logically coherent way. Therefore, the cause of their behavior may not be explainable in a logically coherent way.
If this is true, then you have conceded my point, and we would never have justification for claiming that the same thing causes human behavior and the creation of life. But I don't think that this is true at all. We have valuable explanations for all kinds of aspects of human behavior, from psychology and the neurosciences, even if we are bad at predicting human behavior. Just like the weather.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 02. April 2007 23:29
quote:
My point exactly. Humans (intelligent agents, and so incomparable to weather patterns), to a large degree, do not behave in a logically coherent way. Therefore, the cause of their behavior may not be explainable in a logically coherent way.
If logical coherence isn't a requirement for something to be true, then there's no point in discussing anything. Everything's true and everything's false. Are you sure you want to go that route?
IP: Logged
|
|
LifeEngineer
Member
Member # 3446
|
posted 03. April 2007 09:37
Quote: We understand things scientifically in terms of two types of causes: fixed law, and chance. So what explains inconsistencies in human behavior?
This is inaccurate. Chance or stochastic is a form of a fixed law. Science recognizes deterministic (fixed laws or predictable) and non-deterministic( not describable by fixed laws and scientific theories).
Science or scientific theories only directly address the deterministic component of behavior, but they recognize the indeterminant component of behavior with the 'under ideal conditiions' or 'within defined constraints' qualifiers.
Inconsistencies in human behavior or indeterminant or unpredictable components of behavior would explained or recognized by the qualifiers in predictive theories.
Predictive theories of human behavior have the general form "Under ideal conditions, humans act in a manner compatible with the goal of survival (or the expected or percieved goal of survival)" Departures from this expected or predicted behavior, and there are lots of such deviations, are characterized by science as noise or inefficiency.
IP: Logged
|
|
LifeEngineer
Member
Member # 3446
|
posted 03. April 2007 10:10
Quote: Your suggestion, if I understand it, is to characterize intelligence as "that which is Turing computable". Again, since as far as we know everything is Turing computable, there is nothing that theoretically can't be modelled by a Universal Turing Machine (UTM).
You persist in this misrepresntation despite the fact that I explicitly stated it is incorrect.
As I stated earlier, the logical mathematical framework I use to define both 'intelligent causation' and 'non-intelligent causation' involves deterministic information processing (a Turing machine) and a frame work for classifying output or responses as goal-compatible or non-goal compatible.
Within this mathematical logical framework, we can define the abstract concept of intelligent causation as "If the processing or output of a system changes from time t=0 to time t=1 with the result at time t=1 being goal compatible, and if the change is faster or more efficient than can be explained by a random search, then the change involves intelligent causation. If the criteria is not satisfied, the change does not involve intelligent causation".
In less technical terms, behavior involves intelligent causation if it changes or adapts in a goal directed manner at a rate that is faster or more efficient than random change. Expressed in a still other words, information processing involves intelligent caustion if problems are solved at a pace or with an efficiency greater than trial and error problem solving. This should provide at least a good starting point for a formal scientific definition of intelligent causation.
It should be obvious this definition is applicable to both human behavior and to biological evolution.
The only substantive issues are whether 1) there is an better alternative scientific definition and 2) Is the definition useful in formulating predictive theories.
IP: Logged
|
|
aiguy
Member
Member # 3736
|
posted 03. April 2007 12:20
quote:
As I stated earlier, the logical mathematical framework I use to define both 'intelligent causation' and 'non-intelligent causation' involves deterministic information processing (a Turing machine) and a frame work for classifying output or responses as goal-compatible or non-goal compatible.
Please give me 1) an example of something you believe can be modelled using your model of intelligence, and 2) an example of something that cannot. For (1), show the model.
[ 03. April 2007, 12:21: Message edited by: aiguy ]
IP: Logged
|
|
|
Printer-friendly view of this topic
