On Sun, May 24, 2015 at 12:40 AM, Pierz <[email protected]> wrote: > > > On Sunday, May 24, 2015 at 1:07:15 AM UTC+10, Jason wrote: >> >> >> >> On Tue, May 19, 2015 at 12:44 PM, Bruno Marchal <[email protected]> wrote: >> >>> >>> On 19 May 2015, at 15:53, Jason Resch wrote: >>> >>> >>> >>> On Tue, May 19, 2015 at 12:06 AM, Stathis Papaioannou <[email protected] >>> > wrote: >>> >>>> On 19 May 2015 at 14:45, Jason Resch <[email protected]> wrote: >>>> > >>>> > >>>> > On Mon, May 18, 2015 at 9:21 PM, Stathis Papaioannou < >>>> [email protected]> >>>> > wrote: >>>> >>>> >> >>>> >> On 19 May 2015 at 11:02, Jason Resch <[email protected]> wrote: >>>> >> >>>> >> > I think you're not taking into account the level of the functional >>>> >> > substitution. Of course functionally equivalent silicon and >>>> functionally >>>> >> > equivalent neurons can (under functionalism) both instantiate the >>>> same >>>> >> > consciousness. But a calculator computing 2+3 cannot substitute >>>> for a >>>> >> > human >>>> >> > brain computing 2+3 and produce the same consciousness. >>>> >> >>>> >> In a gradual replacement the substitution must obviously be at a >>>> level >>>> >> sufficient to maintain the function of the whole brain. Sticking a >>>> >> calculator in it won't work. >>>> >> >>>> >> > Do you think a "Blockhead" that was functionally equivalent to you >>>> (it >>>> >> > could >>>> >> > fool all your friends and family in a Turing test scenario into >>>> thinking >>>> >> > it >>>> >> > was intact you) would be conscious in the same way as you? >>>> >> >>>> >> Not necessarily, just as an actor may not be conscious in the same >>>> way >>>> >> as me. But I suspect the Blockhead would be conscious; the intuition >>>> >> that a lookup table can't be conscious is like the intuition that an >>>> >> electric circuit can't be conscious. >>>> >> >>>> > >>>> > I don't see an equivalency between those intuitions. A lookup table >>>> has a >>>> > bounded and very low degree of computational complexity: all answers >>>> to all >>>> > queries are answered in constant time. >>>> > >>>> > While the table itself may have an arbitrarily high information >>>> content, >>>> > what in the software of the lookup table program is there to >>>> > appreciate/understand/know that information? >>>> >>>> Understanding emerges from the fact that the lookup table is immensely >>>> large. It could be wrong, but I don't think it is obviously less >>>> plausible than understanding emerging from a Turing machine made of >>>> tin cans. >>>> >>>> >>>> >>> The lookup table is intelligent or at least offers the appearance of >>> intelligence, but it makes the maximum possible advantage of the space-time >>> trade off: http://en.wikipedia.org/wiki/Space–time_tradeoff >>> >>> The tin-can Turing machine is unbounded in its potential computational >>> complexity, there's no reason to be a bio- or silico-chauvinist against it. >>> However, by definition, a lookup table has near zero computational >>> complexity, no retained state. >>> >>> >>> But it is counterfactually correct on a large range spectrum. Of course, >>> it has to be infinite to be genuinely counterfactual-correct. >>> >>> >> But the structure of the counterfactuals is identical regardless of the >> inputs and outputs in its lookup table. If you replaced all of its outputs >> with random strings, would that change its consciousness? What if there >> existed a special decoding book, which was a one-time-pad that could decode >> its random answers? Would the existence of this book make it more conscious >> than if this book did not exist? If there is zero information content in >> the outputs returned by the lookup table it might as well return all "X" >> characters as its response to any query, but then would any program that >> just returns a string of "X"'s be conscious? >> >> I really like this argument, even though I once came up with a (bad) > attempt to refute it. I wish it received more attention because it does > cast quite a penetrating light on the issue. What you're suggesting is > effectively the cache pattern in computer programming, where we trade > memory resources for computational resources. Instead of repeating a > resource-intensive computation, we store the inputs and outputs for later > regurgitation. >
How is this different from a movie recording of brain activity (which most on the list seem to agree is not conscious)? The lookup table is just a really long recording, only we use the input to determine to which section of the recording to fast-forward/rewind to. > The cached results 'store' intelligence in an analogous way to the storage > of energy as potential energy. We effectively flatten out time (the > computational process) into the spatial dimension (memory). > But what if intelligence were not used to create the lookup table? Does the history/creation event, which may be arbitrarily far in the past, really play any role in a present level of consciousness? Or the reverse: the future creation of a one-time-pad, retroactively restores intelligence (consciousness?) to what was previously always a dumb program. > The cache pattern does not allow us to cheat the law that intelligent work > must be done in order to produce intelligent results, it merely allows us > to do that work at a time that suits us. The intelligence has been > transferred into the spatial relationships built into the table, > intelligent relationships we can only discover by doing the computations. > The lookup table is useless without its index. So what your thought > experiment points out is pretty fascinating: that intelligence can be > manifested spatially as well as temporally, contrary to our common-sense > intuition, and that the intelligence of a machine does not have to be in > real time. That actually supports the MGA if anything - because > computations are abstractions outside of time and space. We should not > forget that the memory resources required to duplicate any kind of > intelligent computer would be absolutely enormous, and the lookup table, > although structurally simple, would embody just a vast amount of > computational intelligence. > > There is a common programming technique called memoization. Essentially building automatic caches for functions within a program. I wonder: would adding memorization to the functions implementing an AI eventually result in it becoming a zombie recording rather than a program, if it were fed all the same inputs a second time? Perhaps the recording of a counterfactual relation retains its effectiveness. Perhaps its impossible to differentiate the two and cases and ascribe consciousness to one but not the other. Jason > > >> A lookup table might have some primitive conscious, but I think any >> consciousness it has would be more or less the same regardless of the >> number of entries within that lookup table. With more entries, its >> information content grows, but it's capacity to process, interpret, or >> understand that information remains constant. >> >> >>> >>> Does an ant trained to perform the look table's operation become more >>> aware when placed in a vast library than when placed on a small bookshelf, >>> to perform the identical function? >>> >>> >>> Are you not doing the Searle's level confusion? >>> >> >> I see the close parallel, but I hope not. The input to the ant when >> interpreted as a binary string is a number, that tells the ant how many >> pages to walk past to get to the page containing the answer, where the ant >> stops the paper is read. I don't see how this system consisting of the ant, >> and the library, is conscious. The system is intelligent, in that it >> provides meaningful answers to queries, but it processes no information >> besides evaluating the magnitude of an input (represented as a number) and >> then jumping to that offset to read that memory location. Can there be >> consciousness in a simple "A implies B" relation? >> >> >>> The consciousness (if there is one) is the consciousness of the person, >>> incarnated in the program. It is not the consciousness of the low level >>> processor, no more than the physicality which supports the ant and the >>> table. >>> >>> Again, with comp there is never any problem with all of this. The >>> consciousness is an immaterial attribute of an immaterial program/machine's >>> soul, which is defined exclusively by a class of true number relations. >>> >>> >> While I can see certain very complex number relations leading to a >> human-level consciousness, I don't find that kind of complexity present in >> the relations defining a lookup table. Especially because any meaning or >> interpretation of the output depends on the person querying it, there's no >> self-contained understanding of the program's own output. >> >> >>> The task of a 3p machine consists only in associating that consciousness >>> to your local reality, but the body of the machine, or whatever 3p you can >>> associate to the machine, is not conscious, and, to be sure, does not even >>> exist as such. >>> >>> I am aware it is hard to swallow, but there is no contradiction (so >>> far). And to keep comp, and avoid attributing a mind or worst a "partial" >>> mind to people without brain, or to movie (which handles only very simple >>> computations (projections)), I don't see any other option (but fake magic). >>> >>> It is perhaps helpful to see that this reversal makes a theory like >>> Robinson arithmetic into a TOE, and to start directly with it. >>> >>> In that case all what we deal with is defined in term of arithmetical >>> formula, that is, in term of 0, s, + and *. >>> >>> The handling of the difference between object and their description is >>> made explicit, through the coding, or Gödel-numbering, or programming, of >>> the object concerned. >>> >>> For example, in the combinators, the number 0 is some times defined by >>> the combinator SKK, the expression "SKK" can be represented by a Gödel >>> number (in many different way), attributing to 0 a rather big number >>> representing its definition, and this distinguish well 0 and its >>> representation (which will be a rather big number). Proceeding in this way, >>> we avoid the easy confusions between the object level and the metalevel, >>> and we can even mixed them in a clean way, as the metalevel embeds itself >>> at the object level (which is what made Gödel and Löb's arithmetical >>> self-reference possible to start with). >>> >>> I would have believed this almost refute comp, if there were not that >>> quantum-Everett confirmation of that admittedly shocking self-diffraction. >>> (which is actually nothing compared to the one with information elimination >>> or dissociation, needed for a "semantic of first person dying" in that >>> self-diffracting reality). >>> >>> We can't know the truth, but we can study the consequence of comp about >>> that, and what machines can say about that, and justify, not justify, >>> expresses or not expresses, hope or fear, etc. >>> >>> >> I appreciated the reversal, I suppose my line of questioning should be >> interpreted as what classes of programs in arithmetic/RA/platonia have the >> capacity to support or add measure to first person views of the type humans >> seem to have. >> >> Jason >> >> -- > You received this message because you are subscribed to the Google Groups > "Everything List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/everything-list. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.
