[ExI] Red

[John Clark]

On Fri, Jul 10, 2026 at 4:39 PM Brent Allsop via extropy-chat <
extropy-chat at lists.extropy.org> wrote:


*>> As I've mentioned before, unless the Corpus Callosum has been
>>>> surgically severed, the two hemispheres of the brain are so intimately
>>>> connected that the left hemisphere does not know what just being the right
>>>> hemisphere would be like, and the right hemisphere doesn't know what just
>>>> being the right hemisphere would be like either. *
>>>>
>>>
>>> *> Yes, we keep repeating this.  And my same old reply is:  I agree with
>>> you at the composite qualia level, where all knowledge is infallibly
>>> subjectively bound into one unified experience.Surely you must admit that
>>> these composite experiences are composed of something at some elemental
>>> level. *
>>>
>>
>> *I do agree, and you can't get more elementary than changing on to off,
>> or off to on. *
>>
>> * > i.e. the painting is composed of pixelated elements representing
>>> particular qualities. *
>>>
>>
>> *A pixel is part of a painting, and a bit is part of a pixel, and a bit
>> is composed of an on/off switch. *
>>
>>
>>> * > You'd have to be the entire painting to know what that painting, in
>>> its entirety, is like. *
>>>
>>
>> *And a Shakespearean sonnet is composed of letters, and the letters are
>> composed of bits, and a bit is composed of an on/off switch. And you'd have
>> to look at all the switches to know what the entire poem was about.  *
>>
>>
>>> *>> I'd  be interested to know if you read the paper that Jason
>>>>>> mentioned: *
>>>>>> *https://www.anthropic.com/research/global-workspace
>>>>>> <https://www.anthropic.com/research/global-workspace> *
>>>>>
>>>>>
>>>>> *>>>>>> Interesting video, yes.  I haven't made it all the way through
>>>>> yet. *
>>>>
>>>>
>>>> *>>>>This is one of the most interesting AI papers I have ever read, it
>>>> doesn't prove that Claude is conscious of course but it comes closer to
>>>> doing so than I would've thought possible. They use a mathematical
>>>> technique that was discovered about 200 years ago called the "Jacobian"; a
>>>> complicated system may be virtually impossible to understand globally but
>>>> if it is "smooth" (no discontinuous jumps) and if you zoom in enough then
>>>> it looks approximately linear, the Jacobian is that approximation.   *
>>>> *An AI like Claude has trillions of electronic neurons with trillions
>>>> of different activation values. When one activation changes even by a tiny
>>>> amount it causes thousands of other activations to change which is why they
>>>> are so hard to figure out. You can think of Claude as a mind with a
>>>> trillion different directions and most of them don't have much effect on
>>>> what the AI ends up saying and doing, but a very few specific directions
>>>> do, and the Anthropic researchers have manage to find those directions by
>>>> using the Jacobian, they call it the "J-space". A small change in the
>>>> J-space makes it much more likely that Claude will talk about a specific
>>>> concept, like Paris or error or** the red qualia. *
>>>> *When you ask Claude an easy question not much happens in the J space
>>>> just as when you're walking you don't think about placing one leg in front
>>>> of the other, but when you ask Claude to solve a hard problem things become
>>>> much more interesting, the J-space sounds like an internal monologue that
>>>> is very much like a human's, or at least like this human's internal
>>>> monologue when I try to solve a difficult problem.*
>>>>
>>>
>>> *>>> Yes, very interesting.  It's definitely a good emulation of
>>> consciousness. *
>>>
>>
>> *>>E**mulation of consciousness**? When a calculator tells you that
>> 2+2=4 is that "4" a real 4 or an emulated 4? Would you need to repeat the
>> calculation yourself with pen and paper to get hold of a real "4"?*
>>
>
> *> We just disagree here.  You are in the It from Bit camp
> <https://canonizer.com/topic/138-It-from-Bit/1-Agreement?is_tree_open=1>
> and I am in the "NO" camp. **If you can describe (let alone demonstrate)
> a way to get a redness, or any "it" from any kind of "bit," I will consider
> my camp falsified.*
>

*I will never be able to falsify your idea and you will never be able to
falsify my idea, that's a fact and regardless of what Karl Popper says** there
is a fundamental reason why my idea is better. Personally I dislike the
idea of brute facts but I dislike an infinite sequence of iterated "how"
questions even more, so I figure  if you must have a brute fact I would
like it to be as simple as possible, but you just ignore the entire matter.
I could be wrong but I don't think you believe the chain of "why" questions
goes on forever so you must believe in brute facts, but if you do I am
quite certain that the thing you think terminates them is a hell of a lot
more complicated than on to off. And you're going to need a lot more brute
facts than I need, so that's why I think my idea is better.   *


> *> The it from bit camp seems unfalsifiable to me.*
>

*It from bit is unfalsifiable, and so is whatever you call your camp,
**Karl Popper
might be spinning in his grave **but that doesn't change the FACT that a
sequence of "how" questions either goes on forever or terminates in a brute
fact.  *

*>>> But it is still not engineered to be like anything like our visual
>>> knowledge is.*
>>>
>>
>> *>> Maybe, but we don't know enough for you to make a definitive
>> statement like that. We know how the AI was engineered (or at least we did
>> before the AI started to reengineer itself) but we know much less about the
>> human brain. *
>>
>
> *> "Maybe"?  We can engineer a brain to represent red with any quality we
> want (like greenness) just by inverting the signal (either before or after
> the retina), or by stimulating the neural cortex correctly. When people
> engineer cortical color neuroprostheses, they will be able to use any color
> qualities they want to represent red.*
>

*I don't see your point. There is no need to resort to brain surgery to
cause somebody to see a strawberry in a color other than red, just change
the lighting conditions.  *


> *> All you can say is "some miracle happens here" when you claim redness
> can come from a bit,*
>

*And you can do better? Of course you can't! Even pure mathematics can NOT
start with nothing, mathematics needs to start with axioms that are simple
and do not seem unreasonable but have no proof. You say a miracle is needed
to turn intelligent data processing into consciousness, but I say it's not
a miracle it's just when you dig deep enough into the fundamental nature of
reality you will eventually encounter a brute fact and one of them is
that* *consciousness
is the way data feels when it is being processed intelligently. After all,
there is no law of logic that demands every event have a cause.*

*So perhaps an even greater mystery is why does ANYTHING have a cause? To
that question I can only say that if the universe was totally random, that
is to say if nothing had a cause, then we wouldn't be here to ask any
questions. Yes I know that's not a good answer but it's the best I can come
up with. *

*You want to find the fundamental mechanism that changes objectivity to
subjectivity and I can sympathize with that desire, I want to do the same
thing. But if you want to do that don't concentrate on objectivity or
subjectivity, concentrate on "mechanism", find the simplest one possible,
and that would be a machine that changes on to off or off to on.   *

*There is no proof AIs are conscious and never will be, but given the
mountain of evidence that keeps increasing each day don't you think it
would be wise if when you encounter an Artificial Intelligence your default
position should be to assume the AI is conscious? After all, I am quite
certain that is already your default position whenever you encounter an
intelligent human being.  *

*When you meet an intelligent entity do you really need to know if its
brain is wet and squishy or dry and hard before you can determine if it's
conscious or not?  *

*John K Clark*



>
>
>>
>>
>>
>>
>>
>>
>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>>
>>>>>
>>>>>> *Yes and that is the fundamental problem you will never be able to
>>>>>> overcome, there is a limit to what cause-and-effect can tell you, and
>>>>>> therefore there is a limit to what science can tell you. Quantum Mechanics
>>>>>> can say why some molecules reflect red light and other molecules reflect
>>>>>> green light but that's not what you're talking about, you're talking about
>>>>>> color qualia. You could maintain it's a brute fact that Glutamate has the
>>>>>> green quality and glycine has the red quality; but both those molecules are
>>>>>> extremely complicated objects and yet they produce an extremely narrow very
>>>>>> specific qualia, so to explain the entire conscious experience with your
>>>>>> theory you're going to need an astronomical number of brute facts, perhaps
>>>>>> even an infinite number of them. But my theory only needs 2 brute facts:*
>>>>>>
>>>>>> *1) On is different from off. *
>>>>>> *2) Consciousness is the way data feels when it is being processed
>>>>>> intelligently. *
>>>>>>
>>>>>> *Therefore William of Ockham would say that my theory is superior to
>>>>>> your theory. *
>>>>>>
>>>>>>
>>>>>>> *> We simply need nature to tell us, through direct apprehension,
>>>>>>> which one is which, so we can have our grounded dictionary of the physical
>>>>>>> qualities of what we are objectively observing.*
>>>>>>>
>>>>>>
>>>>>> *This dictionary you keep talking about is going to be huge, perhaps
>>>>>> going beyond huge and merging with the infinite; even if by some miracle
>>>>>> you possessed such a monstrosity would you really feel that the
>>>>>> consciousness problem had been solved? Wouldn't you want to know WHY one
>>>>>> molecule has the happiness property while another slightly different
>>>>>> molecule has this sadness property? *
>>>>>>
>>>>>> *We're getting to the point where maintaining that AIs are not
>>>>>> conscious is becoming indistinguishable from maintaining that solipsism is
>>>>>> true. *
>>>>>>
>>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260711/97887b51/attachment.htm>