Re: MacLennan: Grounding Analogue Computers

From: Hudson, Joe (jh798@ecs.soton.ac.uk)
Date: Sun Jun 17 2001 - 20:04:47 BST


Hi Stevan,
I see now a lot of my points were not very clearly put. I've attempted to
clarify them.

> > > Hudson:
> > > whatever analogue computation is a digital computer could approximate
> > > the physical behaviour of the implementation to the nth degree
> >
> > >HARNAD:
> > >This just means that a digital computer could simulate any continuous
> > >process to as close an approximation as we wish. But this is still
> > >just simulation (Turing Equivalence, or even Strong Equivalence -- as
> > >close as we like). But, for the same reason that simulated flying (no
> > >matter how closely it approximates it) is not real flying, simulated
> > >continuity is not real continuity.
> >
> > Hudson:
> > Sorry to go through this yet again but what is real flying?

>Harnad
> The stuff that gets you to Singapore. Best not to lose sight of that,
> wherever your thoughts about computation/cognition, analog/digital
> might take you.
>
> (Gets you, the real you, to Singapore, the real Singapore. No
> tricks.)

Hudson
See the summarising points 1/, 2/:

"1/ Implementation independence/dependence (II, ID) depends on your
 plane of awareness, i.e. for TimII the airplane he sees is ID but for
 us (outside all the VR booths) it is II simulation. What is to say we
 are not in the same position as TimII?"

" 2/ We have no basis to say that our plane of awareness is somehow the
 original and 'true' physical existence."

> > Hudson:
> > If the
> > distinction between real and simulated is just that for the real thing
> > the implementation is critical, then we can recreate the same situation
> > inside a VR booth.
>

>Harnad
> No the difference isn't just implementation-independence.
> "Implementation-DEpendence" is simply one of the properties of a
> dynamical (analog) system.

Hudson
What other decisive distinction can always be made between analogue
systems and discrete computational systems? Analogue systems may
also be systematically interpretable so this is not a decisive distinction.

>Harnad
> VR's no good because it doesn't get your (real) body to (real)
> Singapore, it merely fools your (real) senses into feeling as if
> you've gone to Singapore.

Hudson
This is basically the same objection as the one I phrased with the heart
stopping scenario. The heart is just a dramatic way to make the (faulty)
point that where we are physically situated allows us to tell reality from
illusion (or VR simulation).

With the assumption that we are nothing more than a causal physical
system (and lets be honest how is such a system ever going to feel of its
own accord?) we effectively fall into the same category as the airplane,
so called implementation dependent. But from summery points 1/ & 2/,
ID and II just depend on your plane of awareness (I know you don't like
that term but I can't think of a better way to put it.) From inside the VR
system what we touch, smell, hear e.t.c is not II to us, but it is being
created by the II code of the VR sim. The second VR sim inside the first
one is II to us until we step inside the second VR booth. If we had no
experience of the 'real' world we would cite the same difference between
'actually' flying and just being fooled into thinking we were flying in a VR
sim, even though we were actually always in one ourselves (basically
the Matrix scenario).

The functionalist says all that matters is behaviour/functionality.
But all we have to asses this is our senses, hence if while in a VR sim
(whether we knew it or not) we see a 'person' act just like a person usually
does then from a functionalist point of view that is a real person (and a
computational one too). So there is not that much difference between
functionalists and computationalists (of the kind that only aim for Turing
indistinguishability, anyway). Of course even if we are not functionalists
we still have no basis to discriminate (Turing indistinquishability).

The main point here which I should have make explicit before is: If
our senses could be completely fooled (and they are very foolable) then it
only makes sense to talk about ID and II in relative and not absolute terms.

> > Hudson:
> > Now we are inside the fully immersive (caters for
> > all the bodies senses) VR system. We 'see' a plane flying over our head
> > and at the same time a computer on a desk by a VR booth in front of us.
> > Every atom of the computer is fully modelled by the VR simulation. The
> > computer is running a VR simulation of a plane flying.
> >
> > At this point we hastily step outside the VR booth. Now both the VR
> > simulation and its simulation of another VR simulation merge into the
> > same domain of implementation independent computation, as I'm sure you
> > would agree.
>

>Harnad
> Kid-sib doesn't understand all these words. I put on the sensors, I
> see a plane flying. I take them off, I see a computer and VR
> equipment. What's the point. It's obvious what's really going on: A
> computer programme is driving the VR to my senses. It's completely
> irrelevant how well it encodes all the properties of a plane
> symbolically, and then transduces them into inputs for my senses.
> What there is is a computer, a programme it's running
> ("implementation-independently"), VR peripherals, and me (and my
> senses).
>
> > Hudson:
> > However if we step back inside the VR booth, from this
> > perspective ( and that surely is all we have, just a distorted
> > perspective) the plane we see above our heads is not implementation
> > independent.
>
> Kid-sib:
> Nothing of the sort. We don't have a "perspective."
> We have a computer, a programme... etc.
>
> What on earth is a "perspective" supposed to be, in this context?
> The only perspective I know of is (1) the technical stuff of artists
> and architects, and (2) they way things look to me, via my
> eye-balls.
>
> > Hudson:
> > If according to the (hidden) aerodynamics model the wings
> > are the wrong shape then the plane wont fly. Sure we could say that its
> > all really running on a computer and that computer could be any shape
> > or size so long as it was Turing equivalent, but then the computer is
> > no where to be seen, as we are in the VR sim. i.e there is no evidence
> > that we could possibly gather from within the VR sim that the plane we
> > see flying above our head is anything but implementation DEPENDENT, or
> > for that matter a flower we might touch and smell isn't.

>Harnad
> Kid-sib:
> This "implementation dependence/independence" thing seems to
> have become some sort of talisman here. I thought all it meant was
> that it's the programme that matters, not what computer you run it
> on. Whereas with a plane, it's being able to fly that matters, and
> that DOES depend on the device you try to do it with.
>
> And like I said before, we're not talking about tricking human senses
> (VR), but about what's really going on.
>
> (Didn't the course lecturer once warn us about not mixing up "epistemic"
> questions [about what you can KNOW, with your senses, for example]
> with "ontic" questions, about what's really going on? I mean, I hate
> fancy words, but it sounds like that's what's going wrong in your VR
> argument: You're mixing up what's really going on there (computer,
> etc.) with what your senses can tell you under those conditions.)
>

Hudson
And how do we find out what's really going on? Through our senses?
But wait a minute, our senses are rather fallible aren't they?

I have no problem with assuming that what we sense is an accurate portion
of what's really going on, for the purpose of getting on with day to day
life.
But I think to go on and state without reservation that what we sense really
is what's happening is mistaken, even though our senses are all we have to
go on. My above comments are developed on the basis of our being in this
uncertain position.

> > Hudson:
> > Now (still in the VR booth) our attention is drawn to the computer on
> > the desk. We see a plane flying across the screen. Obviously just an
> > implementation independent simulation. We decide against our better
> > judgement to step inside this VR booth. Now by the same token what was
> > implementation independent becomes implementation dependent. We can
> > imagine an infinite regress.
>
> Kid-sib:
> Ya lost me...
>
> > Hudson:
> > Could all this be countered by pointing out that if your heart stops in
> > the real world then all those nested simulations vanish and only the
> > real world remains, hence all they ever were were computer simulations?
> > Not completely.
>
> Kid-sib:
> Was there something in this course that would have made you think
> otherwise? Like that we were immaterial ghosts or something? (I'm not
> even sure it makes sense to say we're "using" are brains: Rather, we
> ARE our brains, or some functional parts of them...)
>

Hudson
Yes. The lack of any clear augment to suggest that we are just our brain.
Also my intuition is not at all happy about the thought. I can't begin to
fathom
how anything which is essentially just a sophisticated clock work mechanism
with senses to regulate its timing, could ever give rise to feeling or
consciousness. Also as you have said there would be no evolutionary
advantage to such things, so how did it come about? To say 'chance' is like
ignoring the question.

> > Hudson:
> > I think it can be excepted that our brains
> > are causal systems and hence there functionality can be recreated in
> > simulation.
>
> Kid-sib:
> Our brains (like everything else, as far as I can tell) are "causal
> systems" alright. And they can be simulated (i.e., a symbol system
> can be designed whose symbol manipulations can be systematically
> interpreted as corresponding to properties of the brain). That's the
> "Church/Turing Thesis" (physical version, CTTP), right?
>
> > Hudson:
> > But could the actual 'physical' implementation of our
> > brains matter somehow? Well if we say it is only our brain which
> > governs our behaviour and we imagine a body with a brain (TimII) in our
> > VR sim then it should act in exactly the same sort of way we do. So
> > from a functionalist point of view we are implementation independent.
>

> > Hudson:
> > This leads to the observation that if TimII also enters the second VR
> > booth and his heart, which was Turing indistinguishable to us from our
> > own in the first VR sim, stops then he will cease to function in the
> > second VR sim too. All this means is that TimII's existence will halt
> > in all VR sim nested levels up to the one where he was first defined
> > and physically modelled. For a functionalist there should be no
> > essential difference between 'physically defined' and 'physically
> > situated'. This objection only illustrates the significance of the
> > 'original VR sim nest level', it does not in essence distinguish
> > between reality and simulation.
>
> Kid-sib:
> I don't know about "functionalists," but to ME is certainly seems you
> are just talking about squiggles here...
>
> > Hudson:
> > 1/ Implementation independence/dependence (II, ID) depends on your
> > plane of awareness, i.e. for TimII the airplane he sees is ID but for
> > us (outside all the VR booths) it is II simulation. What is to say we
> > are not in the same position as TimII?
>
> Kid-sib:
> Sounds like the magic work that "perspective" was doing for you
> before is now being done for you by "plane of awareness" and
> "position" -- and you're still treating implementation
> independence/dependence like some sort of mystic thing. Whereas
> what's really going on in so plain (if you're not being tricked by
> VR): computers, squiggling...
>

Hudson
I really don't see how its plain at all. What do you mean by 'magic work'?

> > Hudson:
> > 2/ We have no basis to say that our plane of awareness is somehow the
> > original and 'true' physical existence.
>
> Kid-sib:
> ??? Sounds vaguely Buddhist...
>

Hudson
Hard to show to be false though I think.

> > Hudson:
> > 3/ When we say 'real flying' we should be aware of 1/ and 2/.
>
> > Hudson:
> > Again symbolic systems are disallowed from being analogue. But why?
> > Did my example (Tom the AI) not show how a symbolic system is best
> > viewed as continuious? I say 'viewed as continuious' and not 'is
> > continuious' because any physicaly implemented system (yes even an
> > airplane) can be viewed as discrete at one scale and continuious at
> > another. Something that is considered analogue (in this sense) just
> > means 'best viewed as' continuious.
>
> Kid-sib:
> "Interpretable as" or "viewed as" are both "epistemic". I'm
> interested in what analog systems ARE ("ontic"), not what they can be
> "viewed as" (whether by my mind, intepreting a computer's squiggles, or
> by my senses, through VR driven by a computer's squiggles).
>

Hudson
A true analogue system is one where ANY descretization would alter the
functionality of the system. But any physical system will appear discrete at
one scale and analogue/continuous at another scale, hence we can not tell
whether a system is intrinsically analogue. So it only makes sense to talk
in
terms of "best viewed as".

>Harnad
> Now there you really have put your finger on a problem (traditionally
> called the "mind/body" problem, but better called the
> "feeling/function" problem). And that one beats the heck out of me...
>
> http://www.cogsci.soton.ac.uk/~harnad/Tp/bookrev.htm

Hudson
Just a thought I've been toying with recently. Could the dualism/telekinesis
answer to the hard problem be neatly integrated with the rest of science at
the scale where strange quantum effects such as particles 'teleporting' and
generally breaking the known laws of physical nature are observed? The
function of the brain would then be to amplify these quantum effects (which
would be controlled by our consciousness) upto electrical impulses and
chemical balances, e.t.c.

> > Hudson:
> > Regarding the idea of a complete world simulation where we could in
> > principle fastforward to see ourselves that you spoke of on a few
> > occasions. In this simulation would be a simulation of this simulation
> > (and so on) would it not? Doesn't this leads to an infinite performance
> > capacity? (e.g. imagine placing the simulation virtual viewing camera
> > at the viewing screen of the simulated simulation). In which case it is
> > not physically realisable.

>Harnad
> So what if it leads to an infinite regress? So does a mirror facing a
> mirror...
>

Hudson
I could have made the point clearer. In the case of two mirrors facing
each other this is not truly infinite regress, the photons will disperse and
the reflected light intensity will eventually drop off to nill.

In the case of our nested simulations, each simulation is a full simulation
of the state of the world (to a level of granularity where all observable
behaviour is recreated), therefore the full simulation needs to be fully
simulated by each of the nested simulations, hence infinite regress. As
the truly infinite number of simulations are running on one physical
implementation, that implementation would need to be infinitely powerful.
Otherwise the computational requirements of the nested simulations
would need to tend to zero which would be absurd.

Joe



This archive was generated by hypermail 2.1.4 : Tue Sep 24 2002 - 18:37:31 BST