Re: The role of logic, & planning ...
Marchal wrote: > > Russell Standish wrote (to George): > > >I don't think Bruno's conclusion is weird. I come to essentially the > >same conclusion in "Occam", without the need for formalising > >"Knowledge", nor the need to use Modal logic. > > The fact that you come to the same conclusion does not mean these > conclusions are not weird. I hope you realise these conclusions run > against the average materialistic aristotelian current scientific > paradigm. Naturally. But then, that's part of its appeal! > > >I would like to think that my exposition is easier to follow than > >Bruno's, but this could simply be a biased viewpoint on my part. I > >welcome comment and criticism on that paper. > > I still believe my general remarks apply to your "why Occam's razor". > (I reprint it and I will reread it once I have more time). > You put to much for me in the hypothesis. Like all physicist you seem > not to be aware of the mind body problem. You are right! What is the mind-body problem? > With comp, what the UDA shows > (and what the graph movie or Maudlin works "proves") is that it is > not possible to attach awareness to worlds or histories. I'm not really sure what you mean by this. In essence, I say that an awareness must experience a history in order to be aware. The Time and Projection postulates of my paper. > The reversal > means really that you need first a theory of consciousness, or a > psychology for deriving the existence of physical beliefs. > I agree that there are similarities in some of our conclusion, but > I am not sure we mean the same by "psychology". I'm reasonably sure we don't, at the level of fine detail. However, that debate can be postponed until other issues have been resolved, such as whether my argument stands up to further scrutiny. > > >Incidently, I didn't mean to imply that this sort of modeling of > >Knowlegde was inappropriate, only that there was no discussion as to > >why one would want to model it in this particular way. > > > The word "model" is tricky. It means different things for logician > and painters (who are using it in the sense of reality) and physicist > and toys builder (who are using it as "theory" or approximation, or > reduction). > Soemtimes I use it in the physicist sense ... > But my approach is more axiomatic. I hope I will be able to give > enough illustrations to help understanding ... > > >Its really the > >same as when Hal Ruhl (and I admit I'm putting words in his mouth > >here, although its consistent with my understanding of his position) > >models the universe by cellular automata. > > Hal Ruhl, like Toffoli, and even like Schmidhuber-2, seems indeed > to search for such "modelisation". > But I do not (and apparently Schmidhuber-1 don't do it either). > > The UD does NOT depend on the choice of a particular formal systems. > The UDA really shows that my "awareness" will be linked with all > implementation of my computationnal extension. > By implementation here I just mean the giving of a program and its > relative UTM interpretation. > > And the provability logics (G and G*) is correct and complete for ALL > sound > classical Universal Machines. In that sense there is no modelisation > at all. And comp is not the hypothesis that my brain can be modelised > by a Turing Machine, it is the act of faith of telling "yes" to the > (mad) surgeon. > > >I notice Bruno has posted a more detailed discourse on this issue, > >which I will digest in due course. > > It is an important one, but it will be fully clear only after > I explain Godel and Lob theorem with enough rigor. > > >Perhaps all he was doing was > >assuming a cultural background of philosophy I have not been exposed > >to. Just as an example, he says most philosophers would agree that > >[]A->A, where []A is interpreted as knowing A. This is clearly a > >different meaning of the word "to know" that we use here in > >Australia. I know of plenty of people who know that God exists. And I > >know of a number of other people who know that God doesn't exist. So, > >by this application of Modal logic, we can conclude that God both > >exists and doesn't exist at the same time, which seems kind of > >illogical. > > To say the least. I must say that I am quite astonished that > Australian can "know" falsities. What is the difference between > knowledge and belief for an Australian ? > A matter of degree, as far as I can tell. > >Perhaps the way out of this mess is to say that I'me really talking > >about belief, ... > > Yes, I think indeed you were talking of "belief". The nice thing > with axiomatic approach is that we will "define" knowledge or > knowability by axiom like K, T, 4. Except that formal provability > will be defined in arithmetic and then we will look at which > formula it obeys. And It does not obeys to knowledge axiom (see > below). > > >...rather than knowledge, however that would imply that > >knowledge is devoid of meaning, since it is impossibl
Re: Belief & Knowledge
On Wed, 2 May 2001, Brent Meeker wrote: > A true belief that has a casual connection with the fact that makes it > true. Knowledge is when predicted.
Belief & Knowledge
On 02-May-01, rwas rwas wrote: >> Just as an example, he says most philosophers >> would agree that >> []A->A, where []A is interpreted as knowing A. This >> is clearly a >> different meaning of the word "to know" that we use >> here in >> Australia. > > I get the impression folks here assume that when one > person knows something, that only that person knows > that something. For other people to know the same > something, they have to discover and assimilate it for > themselves. It also seems that folks here assume > knowledge is some kind of pattern that exists separate > from the truth of surrounding it's existence. > >> From a mystic standpoint, this can't be. To know > something is closer to the analogy of a subscriber > line. When one *knows* something, anything, they > subscribe this pattern. Correct me if I'm wrong, but the usual definition of knowledge is: A true belief that has a casual connection with the fact that makes it true. The standard example is that I may believe that Tom has bought a blue car because I saw him drive up in it. And Tom has bought a blue car - so the belief is true. But it isn't knowledge because the car I saw him drive up in is a rental car, not the one he bought. So in this example there is no casual connection between my belief and the fact that Tom bought a blue care, and hence my true belief is not knowledge. Brent Meeker
Re: The role of logic, & planning ...
> > Just as an example, he says most philosophers > would agree that > []A->A, where []A is interpreted as knowing A. This > is clearly a > different meaning of the word "to know" that we use > here in > Australia. I get the impression folks here assume that when one person knows something, that only that person knows that something. For other people to know the same something, they have to discover and assimilate it for themselves. It also seems that folks here assume knowledge is some kind of pattern that exists separate from the truth of surrounding it's existence. >From a mystic standpoint, this can't be. To know something is closer to the analogy of a subscriber line. When one *knows* something, anything, they subscribe this pattern. Another issue is how folks seem to thing knowledge is inanimate until someone acts on it, like words on a paper being meaningless until someone read them. From a mystic standpoint, that isn't so. Knowledge and expression is simply manifest from one place to another. The knowledge itself is not constrained to the limits of those that would interpret it. Those entities interpret and then express that understanding wherever they happen to be existing. For someone to try to form the basis for existence based on what one thinks others can know in terms of what I've tried to counter, I feel intuitively that they would fail, or not succeed completely. One analogy to explain this is someone caught in an event horizon of a black-hole. The realm formed by this Event-Horizon can be vast, but is still by definition, limited. I see people trying to define existence by illusionary data like someone trying to understand the universe by what he can see from his vantage point in the Event-Horizon. Drawn out, it would look like someone walking in a circle. Eventually, he'd come back to where he started. He might vary his path slightly to see different things, but he'd simply be make the circle bigger. He can never know what lay outside the circle with his given modus operandi. >From my perspective, true knowing, is being what you know. Which implies a great deal on what is truly knowable. If you look at what we're used to here, we have belief and knowing implicitly understood in statistical terms. We know we can walk, we've done it so often, so we don't doubt we can. Belief seems to be predicated on the existence of doubt. True knowing has no constraints of doubt. To know is to be one with that knowledge. This from a mystic standpoint, is true faith. Faith is *not* belief. Faith is knowing. I know of plenty of people who know that > God exists. And I > know of a number of other people who know that God > doesn't exist. So, > by this application of Modal logic, we can conclude > that God both > exists and doesn't exist at the same time, which > seems kind of illogical. > > Perhaps the way out of this mess is to say that I'me > really talking > about belief, rather than knowledge, however that > would imply that > knowledge is devoid of meaning, since it is > impossible to establish > with certainty whether any particular fact is true. > Even Mathematical > proof is contingent upon belief of the efficacy of > the formal proof, Again, I had thought the point of these threads were to try to describe consciousness with the idea in mind of trying to synthesize consciousness in software or some other artificial means. I propose the best way to do this is to know what one is after specifically, then solve the problem of achieving it. If one attempts to use a limited thinking style to implement something interpreted with that same thinking style, the end result would seem to necessarily be limited to perceptional constraints of that thinking style. I get the intuitive sense, that linear or sequential thinking will not result in the kind of achievement we're talking about. Robert W. __ Do You Yahoo!? Yahoo! Auctions - buy the things you want at great prices http://auctions.yahoo.com/
Program for UD
Has anyone proposed a specific implementation for the Universal Dovetailer (UD)? This is a program which runs all possible programs, a little bit at a time, making progress in all of them. For something close, here is Greg Chaitin's program to calculate Omega, the probability that a random program will halt. It comes from from http://www.cs.auckland.ac.nz/CDMTCS/chaitin/omega.l, and is written in his dialect of Lisp. Omega in the limit from below! define (all-bit-strings-of-size k) if = 0 k '(()) (extend-by-one-bit (all-bit-strings-of-size - k 1)) define (extend-by-one-bit x) if atom x nil cons append car x '(0) cons append car x '(1) (extend-by-one-bit cdr x) define (count-halt p) if atom p 0 + if = success car try t 'eval read-exp car p 1 0 (count-halt cdr p) define (omega t) cons (count-halt (all-bit-strings-of-size t)) cons / cons ^ 2 t nil Examples of calling it: (omega 0) (omega 1) (omega 2) (omega 3) (omega 8) To read it, keep in mind that Lisp is a prefix style language, so that the syntax is "operator operands". Also, the single quote means that the following argument is quoted rather than evaluated. The built-in functions car and cdr return the 1st element of a list and the remainder of the list, respectively, and cons puts car and cdr back together to form the original list. The first two functions just return a list of all bit strings of size k. These will be the programs that run. The first function tests if k = 0 and returns (()), otherwise it calls itself on k-1 to get all k-1 bit strings, then calls extend-by-one-bit. The latter takes the first element (car of x) and appends both 0 and 1 to it. Then it recurses on the remainder of the list. So calling all-bit-strings-of-size-k with 1 gives ( (0) (1) ), with 2 gives ( (0 0) (0 1) (1 0) (1 1) ) and so on. These are the possible programs which will be run. The count-halt function will return the number of programs in list p which halt within t steps. If p is an atom (not a list) it returns 0, else it tries running car p (the first element of p) for t steps and counts 1 or 0 based on halt/no-halt. It recurses on the remainder of p and adds that result to the 1 or 0. The key to this function is Chaitin's operator "try", which takes a number of steps and a program. It runs the program for that many steps and returns a success/still-running flag, plus the output from the program if any. Above Chaitin is only using the success flag to count whether the program has halted. Last we have omega itself, which for parameter t calls count-halt on all strings of length t. It then shows that this needs to be divided by 2^t to get the halting probability. (BTW Chaitin has actual Lisp interpreters which can run this program at http://www.cs.auckland.ac.nz/CDMTCS/chaitin/lm.html.) You could get the effect of a UD, then, by calling omega with successively larger numbers. It would run all 1-bit programs for 1 step, then all 2-bit programs for 2 steps, all 3-bit programs for 3 steps, and so on. It might appear that this will not, for example, run 2-bit programs for 10 steps. However Chaitin's programs are self-delimiting. When you have all 10-bit programs, some of those are 2-bit programs with 8 ignored bits at the end. So in running N-bit programs for N steps, we are also running all K bit programs for N steps, where K < N. This way of doing a UD is wasteful in that we keep restarting each program from the beginning. I think in most conceptions of the UD we assume that each program's state is retained, so that when its turn comes up again, it continues from where it left off. However I think it would be difficult to manage the storage space for this to work. Doing it Chaitin's way might appear to change the frequency with which a program ones so that it departs from the universal measure. If we have two programs of length K and L, where K << L but both are large, it should be that the first program gets running time 2^(L-K) greater than the second program. However program K actually gets in addition L-K runs before we even start running program L, as we build up to programs of size L. In the end this should not matter though as this constant factor will decrease in importance as the size of programs approaches infinity. Running programs of size N >> L >> K, K will get running time 2^(L-K) more than L due to its smaller size, which corresponds to the universal measure. This leads to three questions: - Would all UD programs would correspond to the universal measure, asymptotically? - Is there a way to retain state for all the programs so that you don't have to start over from the beginning? (Although Chaitin's way may be simpler, and if it gives the same probability distribution then we couldn't tell the difference). - Is it necessary
Re: The role of logic, & planning ...
Russell Standish wrote (to George): >I don't think Bruno's conclusion is weird. I come to essentially the >same conclusion in "Occam", without the need for formalising >"Knowledge", nor the need to use Modal logic. The fact that you come to the same conclusion does not mean these conclusions are not weird. I hope you realise these conclusions run against the average materialistic aristotelian current scientific paradigm. >I would like to think that my exposition is easier to follow than >Bruno's, but this could simply be a biased viewpoint on my part. I >welcome comment and criticism on that paper. I still believe my general remarks apply to your "why Occam's razor". (I reprint it and I will reread it once I have more time). You put to much for me in the hypothesis. Like all physicist you seem not to be aware of the mind body problem. With comp, what the UDA shows (and what the graph movie or Maudlin works "proves") is that it is not possible to attach awareness to worlds or histories. The reversal means really that you need first a theory of consciousness, or a psychology for deriving the existence of physical beliefs. I agree that there are similarities in some of our conclusion, but I am not sure we mean the same by "psychology". >Incidently, I didn't mean to imply that this sort of modeling of >Knowlegde was inappropriate, only that there was no discussion as to >why one would want to model it in this particular way. The word "model" is tricky. It means different things for logician and painters (who are using it in the sense of reality) and physicist and toys builder (who are using it as "theory" or approximation, or reduction). Soemtimes I use it in the physicist sense ... But my approach is more axiomatic. I hope I will be able to give enough illustrations to help understanding ... >Its really the >same as when Hal Ruhl (and I admit I'm putting words in his mouth >here, although its consistent with my understanding of his position) >models the universe by cellular automata. Hal Ruhl, like Toffoli, and even like Schmidhuber-2, seems indeed to search for such "modelisation". But I do not (and apparently Schmidhuber-1 don't do it either). The UD does NOT depend on the choice of a particular formal systems. The UDA really shows that my "awareness" will be linked with all implementation of my computationnal extension. By implementation here I just mean the giving of a program and its relative UTM interpretation. And the provability logics (G and G*) is correct and complete for ALL sound classical Universal Machines. In that sense there is no modelisation at all. And comp is not the hypothesis that my brain can be modelised by a Turing Machine, it is the act of faith of telling "yes" to the (mad) surgeon. >I notice Bruno has posted a more detailed discourse on this issue, >which I will digest in due course. It is an important one, but it will be fully clear only after I explain Godel and Lob theorem with enough rigor. >Perhaps all he was doing was >assuming a cultural background of philosophy I have not been exposed >to. Just as an example, he says most philosophers would agree that >[]A->A, where []A is interpreted as knowing A. This is clearly a >different meaning of the word "to know" that we use here in >Australia. I know of plenty of people who know that God exists. And I >know of a number of other people who know that God doesn't exist. So, >by this application of Modal logic, we can conclude that God both >exists and doesn't exist at the same time, which seems kind of >illogical. To say the least. I must say that I am quite astonished that Australian can "know" falsities. What is the difference between knowledge and belief for an Australian ? >Perhaps the way out of this mess is to say that I'me really talking >about belief, ... Yes, I think indeed you were talking of "belief". The nice thing with axiomatic approach is that we will "define" knowledge or knowability by axiom like K, T, 4. Except that formal provability will be defined in arithmetic and then we will look at which formula it obeys. And It does not obeys to knowledge axiom (see below). >...rather than knowledge, however that would imply that >knowledge is devoid of meaning, since it is impossible to establish >with certainty whether any particular fact is true. But, at least for a non intuitionnist, or a non constructivist, a proposition could be true independently of our belief or knowledge. A platonist (as I am for arithmetics) has no problem with that. Of course the nuance between truth, provable, believable, knowable, .. are subtil. The crazy thing is that Godel (Lob Solovay) will eventually put an immense light on those nuance. In metaphysics the "royal argument" for explaining that indeed we cannot distinguish knowledge with belief is the dream argument. When we are awake, we cannot know for sure that we are not dreaming. Socrate uses it in his reply to Thaetetus. Descartes, Berkeley and almost all
Re: The role of logic, & planning ...
George wrote:: (Tuesday, May 01, 2001 3:40 PM): > If logic (or some form of logical model) is not the vehicle for describing > reality, then what is? > > George > Describing what? can we "think" of anything beyond the OM (including evtl. 'memories' in it, similarly as momentary mind-content)? Do we (in-mind) have knowledge (- what is it?) beyond that what is IN MIND? now if we call the mechanism of mind a "logic", do we have anything else to use for a thing named "reality", which exists for us only in our mind ? (And let me skip the question: "who's mind?") Do we have any clue to distinguish the OM-stuff from our mindcontent? Consequent arguing curls back to solipsism, no matter how revolting it is. So, while subscribing to George's sentence - in total, I have to question the use of the word 'reality'. Reality is MY virtual interpretation of thoughts about - what I believe - is the world "outside". So are the "objective" instrumental measurements with their explanations. As Dr. Johnson said: the only reality is a stone in my shoe, because it hurts. Does it really, or do I only think it is? John Mikes <[EMAIL PROTECTED]> "http://pages.prodigy.net/jamikes";