On Wed, May 6, 2015 at 12:32 PM, John Clark <[email protected]> wrote:
> Russell Standish wrote: > > http://www.hpcoders.com.au/blog/wp-content/uploads/2014/08/mgaRevisted.pdf > > > The Movie Graph Argument seeks to parlay this into an absurdity, where >> there is no active physical difference between a conscious computation, and >> the mindless replaying of a recording. > > > It seems to me that the first thing to do when starting a Reductio ad > absurdum proof is to make sure the conclusion really is absurd, and this > one isn't > It may not be absurd, but its counter to the assumptions of the experiment, that consciousness is the result of a computation. A recording may be stored statically on a hard drive or other medium, it is not a computation. Nor is it a computation if its smeared out over the dimension of time, which is itself static in the relativistic sense of block time. > > If 2 identical computers are running the same AI program computationalism > says there is only one conscious being. If one of the 2 computers is > destroyed computationalism says there is no subjective difference and > nobody died. Time is only meaningful in its relationship to something else > so if the computer is stopped for a billion years and then started up again > there is no subjective difference. If the computer is reset and the same AI > program is run again computationalism says there is no subjective > difference. If the exact same program is rerun on the computer a trillion > times computationalism says there is still only one conscious experience, > that is to say subjectively things would be exactly the same if it had been > run only one time. When a recording of consciousness is played back does > the consciousness exist during the playback or just when the computer was > actually making calculations? If computationalism is true, and I think it > is, then the answer to that question doesn't make any subjective difference > whatsoever. > > What about intelligence, when somebody in the 21th century reads what > Newton accomplished did the intelligence occur in the 17th century or the > 21th? Operationally it makes no difference. Does this mean that > consciousness and intelligence exists outside the normal bounds of space > and time? In a way it sorta does which is why I'm reluctant to say physics > is definitely more fundamental than mathematics; but in another way it > sorta doesn't because the entire argument hinges on the existence of a > computer that operates according to the laws of physics, which is why I'm > reluctant to say mathematics is definitely more fundamental than physics. I > just don't know. . > > >> > One of the consequences of the universal dovetailer argument is that >> you cannot tell which computer program is you. > > Well that sure didn't take long, already on page 1 we run into one of > those damn ambiguous personal pronouns. And in your scenario there is only > one serial program and it is running the entire multiverse so it would be > very easy indeed to say what program is "you" assuming that personal > pronoun means anything. > >> > For every program that instantiates your current conscious state, there >> are an infinite number of possible continuations of that program, >> corresponding to different possible futures. This leads to an irreduciable >> indeterminism > > There is nothing new in this, Turing proved 80 years ago that in general > the behavior of even very simple programs can not be predicted, if you want > to know what it will do all you can do is watch. And Og the caveman > discovered that people don't always know what they will see next so I see > no reason for a new homemade acronym like "FPI". > >> > even an omniscient god cannot know what you will experience next. > > Because even a omniscient god cannot answer a question that isn’t really a > question but is gibberish. > I thought you were a believer in the logical possibility of fundamental randomness. > > the Church-Turing thesis implies that it doesn’t matter what >> physical computer the dovetailer is run on. > > I'm glad you said "physical", the dovetailer could be made many different > ways but however its made as far as we know it must be constructed of > matter and it must operate according to the laws of physics. > >> >The wrinkle is to suppose that the universe doesn’t have sufficient >> resources to run a universal dovetailer. Whilst a universal dovetailer can >> be coded and run on the physical computers we have today, in practice only >> a short initial portion of the UD can be run. What if the universe goes >> into a heat death before any conscious program is started? To distinguish >> between these cases, Marchal calls a universe capable of running a >> universal dovetailer fully a robust universe. > > > So at the very start you've got to assume the existence of a physical > universe and a very special type of physical universe. Does this mean > physics is more fundamental than mathematics? > This is an intermediate step. The next step dispenses with the need for a physically concrete Dovetailer implementation. > > > we can, for the purposes of this argument, consider robust universes to >> be ones that can run enough of the universal dovetailer for >> programs instantiating all possible human experiences of consciousnesses >> within a human life time be executed. This is still an immense universe, >> but no longer an infinite one. > > > You said the dovetailer "leads to an irreduciable indeterminism", but if > the machine is finite then a faster but still finite computer could predict > what the dovetailer will do; it still could not of course predict what > "you" will see next because in this context that personal pronoun has no > meaning, it would be like asking what klogknee will fluxanate next. > > > Since all our possibe experiences will be instantiated, and observed, >> our phenomenal physics depends only on the properties of the universal >> machine, not on any underlying physical sustrate. > > > But there is one physical substrate that is still of critical importance, > the substrate out of which your dovetailer machine is built because nobody > knows how to make a Turing Machine or a computer or how to make one single > calculation without using matter that operates according to the laws of > physics. Maybe there is some other way to do it but if there is nobody > knows what it is. > > > it is observed that different brain states invariably correspond to >> different conscious experiences. > > > And different brain states lead to different physical behavior. > > >> > Now consider the scenario of a class of school children, one of whom >> is named Alice, and another Bob. Does Alice’s consciousness supervene on >> the class? Well, yes, as we observe that any change in Alice’s >> consciousness must correspond to a physical change in the classroom, >> concentrated in Alice’s brain. > > > Or to say the same thing more simply, the Turing Test (a behavioral test) > works for both intelligence and consciousness. > > > But we can ask a slight different question — does consciousness >> supervene on the class. In this case, we’d have to answer no, because both >> Alice’s conscious states and Bob’s, not to mention the teacher’s and other >> students are all present in the class. A difference in conscious state does >> not correspond to a physical difference. > > > I have no idea what you're talking about here, if one person effects the > class the other class members will effect it too. > > > We can express the same conundrum using the speech case, exploiting the >> so-called “cocktail party” effect. Alice says “hello”, and Bob says >> “hi” simultaneously — but which word we hear depends on who we’re actively >> listening to. > > And the programming of the dovetailer determines who we're actively > listening to. I really don't see what you're drivin at. > >> > The words no longer supervene on the air molecules, but on the >> state of the listener. > > > You've lost me. > > > if program A executes the ”or” instruction on registers x and y, and B >> executes the ”and” instruction. If it so happens that both x and y both >> contain the same value (both true or both false), then the resultant >> machine state is identical with each program. Yet the two programs are >> quite different, as if the two registers had different values, the >> resulting machine state would be quite different. We call this “if it had >> been different” a counterfactual. > > > If quantum mechanics has taught us anything it's that counterfactuals are > a intellectual dead end, you can't make a measurement you didn't make. You > can choose to measure the position of a particle with as much accuracy as > you want but if you do you will have little idea what its momentum would be > and it would be pointless to speculate what the momentum value would have > been if you had chosen to measure that instead of position. > Is Newcomb's paradox meaningless because it concerns counterfactuals? > > > In this case, programs A and B are not counterfactually >> equivalent. It seems plausible that counterfactual inequivalence is needed >> as part of the definition of what could differ between computations >> supporting different conscious experiences. > > > That doesn't seem the least bit plausible to me, as long as the inputs > are both true (or both false) then I don't see how on earth it could make > any difference, the "and" instruction produces the same result as the "or" > instruction and the machine is in the identical state. > Here's an example: Ask a man to compute 2 + 3, and ask a calculator to compute 2 + 3. They both have the same output: 5, but very different intermediate conscious states, > Therefore if computationalism is true then your above statement is more > than just implausible it is flat out wrong. > > > if counterfactual situations are physically realised somewhere in >> the multiverse, then [...] > > > Then if things were different then things would be different. If one > input were 0 and the other 1 then there would be a difference between that > universe and the universe where both inputs were 0 or both 1 and so the > universe would split and the consciousness of the AI program would be > different too so it also would split. > > > This is supported by the intuition that a mere playback of a recording >> of a conscious machine (eg reanimating a dead brain by passing recorded EEG >> sgnals through the neurons) is not sufficient to instantiate >> a consciousness. > > > That isn't my intuition! Audio and video can be recorded and I see no > reason why in theory consciousness couldn't be recorded too, > Because a recording isn't a computation, and computationalism was assumed. It's the difference between Schmidhuber's program that is an infinite loop incrementing and printing i, and Bruno's UDA which treats each i as a program, and executes it as such. Jason > as I explained above if computationalism is true things like that must be > possible, after all... that's what computationalism means. > > >> > The idea is that the conscious computation is implemented as a graph >> (or network) of stateful objects (eg abstract neurons) embedded in a glass >> plate. This allows a movie camera to record a movie of the operation of the >> artificial brain. Then by parts, he severs some of the >> network links between neurons, but by projecting the movie back onto the >> network, is able to excite those neurons as though they were still >> connected. > > > But the neurons still are connected just in a more roundabout way, the > excitation signal from neuron X is recorded on film and then sent to neuron > Y my means of the projector. I really don't see the point of this thought > experiment. > > >> > The requirement that the computations be counterfactually correct >> means two distinct conscious experiences may produce the same sequence of >> machine states, but they must differ in their behaviour for some >> counterfactual machine state. > > > That certainly isn't computationalism and it isn't science either because > science deals in the experiments you made not the experiments you didn't > make. > > John K Clark > > > > -- > 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.
