Re: Rationalizing is something we *all* do

On 02 Nov 2012, at 23:02, Rami Rustom wrote:

> On Nov 1, 2012 1:10 PM, "Bruno Marchal" <[email protected]> wrote:
>  <snip>
> > >
> > Hmm.... OK. Can be quite implicit though, and not recognizable as  
> such. [subconscious idea]
>
> What does that mean?
>

That there are no algorithm to decide what a piece of code or matter
can compute or do.





>
> >
> >
> >
> > > Why would it be off limits from being discovered?
> > >
> > There is no algorithm to decode what a program can do, from its code
> > (Rice theorem). The set of programs doing a particular task is not
> > recursive. With neural nets, it is worst, as the information can be
> > distributed in tiny difference of excitation level of billions of
> > neurons.
>
> When speaking of software, one does not need to speak of hardware. The
> hardware doesn't affect the software, unless the hardware gets
> damaged. Agreed?
>

Yes. But Rice theorem concerns both software, and possible hardware.




>
>
> > >
> > If we are machine, we cannot know which machine we are. We can only
> > bet on some level of description, in case we accept a digital brain
> > prosthesis, for example. Even if the guess is correct, we still  
> cannot
> > recognize the genuine program which constitutes our identity. The
> > relevance of this is that for the gut feelings, we might be unable  
> to
> > decide if some idea is a prejudice of the parents, or of the  
> mammals,
> > or of all universal machines. In the mater case, we can't change it
> > without loosing consistency or soundness.
>
> I still don't know what that means nor why its relevant to our
> discussion. Consistency of what? Soundness of what?
>


Of us, assuming we are machine. All machine looking inward discover
that they have an irrational part, in the sense that their knowledge
grows bigger than what they can prove. That is also why the will
change, in a way that they cannot predict.





>
>
> > >
> > > I don't understand your idea that some subconscious ideas are off
> > > limits.
> > >
> > Intuitively it seems obvious, if only because our brain have  
> plausibly
> > a very long complex history, and nature does not make explicit
> > programs with readable comment.
>
> You are using the theory of the mind that says that certain brain
> parts cause certain mind parts. But that theory has been refuted by
> DD. That refutes evolutionary biology (which is the field you're
> referring to).
>

Where did I talk on brain parts? The conscious mind is not even
related to the brain, which in fine is a construct of the mind. Now if
you use mind in a large sense, then yes, the part of the brain or of
the computer which clean up this of that memory can be attributed to
some part of the computer. If not, your point above that if the
hardware is damaged it has consequence on the mind, would not follow.




>
> What you're saying is that some of our software is hardcoded. And that
> hardcoding evolved over millions of years. But, humans don't have any
> hardcoding. Its all softcoded. All of it changes.
>

This does not make sense for me (despite comp implies non materialism,
but only globally). At some level the brain changes itself, but
through the material behavior of neurons. Then some part of the brain
are more hardcoded than others, like the cerebral stem, the limbic
system. But even the cortical system has an hard coded part, and the
change of itself will be hard changes.




>
> > Then, although machine can represent
> > itself entirely, it can't represent its behavior entirely, almost  
> for
> > reason akin to the reason that nobody can see directlmy his/her own
> > back. there are blind spot in self-reference, for logical reasons.
>
> Do you agree that we can guess about them?
>

About some of them, yes. I doubt we can guess all of them.




>
> >
> >
> >
> >
> > >
> > > Consider this. Decades from now we create the first AGI. It  
> programs
> > > its own code. No?
> > >
> > This is ambiguous. Programs can easily modify, even completely their
> > own code, but it cannot create one from nihilo.
>
> Whats nihilo?
>
Nothing. You need the program to start with, and some contexts.



>
> > There is always
> > another universal program or reality needed, even if it is the
> > arithmetical reality, or physical reality, which are highly
> > undecidable set (probably in the first case, plausibly in the second
> > case).
> > It is like our wanting. We have few control about what we want,
>
> False. Our wants depend on our values. We have control over our
> values. Our values are ideas. We guess ideas and criticism them.
> Sometimes we're doing this with our value-type ideas. And when we make
> changes to our values, we've changed our wants.
>

I have to sneeze now. Oh, I will change that, so I can continue the  
talk.
I want to be happy? Oh, I will change that, life will be so much
simpler.
I am afraid to die. Oh I will change that, so that I can kill myself
in peace.
...
I am not sure that our values are only ideas. That seems a bit too
much constructivist for my appreciation of the comp hypothesis.



>
> > nor do
> > we have control on the way events can hurt us.
>
> Physical hurt or mental hurt? We have a lot of control over mental
> hurt.
>

Hurt is always mental. "physical hurt" is a manner of speaking, of
mental hurt related to low level sensation coming from the periperical
(hardcoded too) nervous system (including glial cells).




> Mental hurt is TCS-coercion. TCS-coercion occurs when someone is
> acting on a theory while another conflicting theory is active in his
> mind. An example of TCS-coercion is someone gets offended when someone
> else says a racial slur to him. So the one who gets hurt doesn't want
> to be called by that racial slur but it happened. Do you think he has
> no control over getting hurt by this?
>

Here he has control. But that's a special case. And then he has
control but he need some education, so even if that is in principle
controllable, most of the time in "real-life" he will not.




>
> > In fact we control very
> > few things, but yet, we do have a non negligible partial control,  
> but
> > it is only a window, so to speak.
>
> It is a window. But, that window can be made bigger without limit. Or,
> is the limit 100%? Can someone be absolutely conscious of *all* his
> ideas such that none of them are subconscious/inexplicit?
>

No, it cannot. In fact consciousness needs some unconsciousness. You
can make a machine referring to 100% of its body (this is not trivial
to prove), but you can't make a program referring to all its possible
behavior.
This has been intuited by Hofstadter, and proved by Solovay, and others.




> Doesn't that
> imply perfection?
>
> No it doesn't. Because he could be wrong about any of them.
>

Not on all of them. There is a fixed point which cannot be doubted,
like consciousness itself.

Bruno

http://iridia.ulb.ac.be/~marchal/





[Non-text portions of this message have been removed]