[ExI] ai emotions

[Stuart LaForge]

Quoting Brent Allsop:


>> ?Consciousness is not magic, it is math.?
>
> How do you get a specific, qualitative definition of the word ?red? from
> any math?

Red is a subset of the set of colors an unaugmented human can see.  
There I just defined it for you mathematically. In math symbols, it  
looks something very much like {red} C {red, orange, yellow, green,  
blue, indigo, violet}. If you were a lucky mutant (or AI) that could  
perceive grue, then the math would look like {red} C {red, orange,  
yellow, green, grue, blue, indigo, violet}

Whatever unique qualia your brain may have assigned to it is your  
business and your business alone since you cannot express red to me  
except by quantitative measure (650 nm wavelength electromagnetic  
wave) or qualitative example (the color of ripe strawberries). Any  
other description of red only means anything to YOU (Perhaps it makes  
your dick hard, I have no clue, don't really care.)


In other words, you can't give me any better a qualitative description  
of red then I can give you. Prove me wrong: What is red, oh privileged  
seer of qualia?  (Yes, that was sarcasm.)

>
> ?I don't think that substrate-specific details matter that much.?
>
> Then you are not talking about consciousness, at all.  You are just talking
> about intelligence.  Consciousness is computationally bound elemental
> qualities, for which there is something, qualitative, for which it is like.

Intelligence and consciousness differ by degree, not by type. Both are  
emergent properties of some configurations of matter. If I were to  
quantitatively rank emergent properties by their PHI value, then I  
would have a distribution as follows: reactivity <= life <=  
intelligence <= consciousness <= civilization

>> ?It is irrelevant that I perceive red as green.?

> Can you not see how sloppy language like this is?  I?m going to describe at
> least two very different possible interpretations of this statement.  If
> you can?t distinguish between them, with your language, then again, you are
> not talking about consciousness:

You pull a single sentence of mine out of context and then use it to  
accuse me of sloppy language? Here is my precise and unequivocal  
retort: NO! I challenge you to take that out of context.

> 1.       One person is color blind, and represents both red things and
> green things with knowledge that has the same physical redness quality.  In
> other words, he is red green color blind.
>
> 2.       One person is qualitatively inverted from the other.  He uses the
> other?s greenness to represent red with and visa versa for green things.

When you said, "Are you talking about your redness, or my redness  
which is like your [sic] grenness?" I meant whichever you meant by the  
quoted statement. My argument holds either way. Unless you believe  
that color-blind people are not really conscious. In which case you  
should be enslaving the colorblind and tithing me 10% of the proceeds.

>
> You can?t tell which one you?re statement is talking about.  Again, you?re
> not talking about consciousness, if you can?t distinguish between these
> types of things with your models and language.

Again, my statement reflects yours with the exact same scope. So you  
tell me what I meant.

> Sure, before Galileo, it didn?t matter if you used a geocentric model of
> the solar system or a heliocentric.  But now that we?re flying up in the
> heavens, one works, and one does not.  Similarly, now, you can claim that
> the qualitative nature doesn?t matter, but as soon as you start hacking the
> brain, amplifying intelligence, connecting multiple brains (like two brain
> hemispheres can be connect) or even religiously predicting what ?spirits?
> and future consciousness will be possible.  One model works, the other does
> not.

I don't see how your model predicts anything except for your ignorance  
of what consciousness is. You say that every consciousness is a unique  
snowflake of amassed qualia, I say that every machine-learning  
algorithm starts out with a random set of parameters and through  
learning its training data, either supervised or unsupervised,  
converges on an approximation of the truth

Every deep learning neural network is a unique snowflake that gets  
optimized for a specific purpose. Some neural networks train very  
quickly, others never quite get what you are trying to teach it. There  
is very much a ghost in the machine and each time you run the  
algorithm, you get a different ghost. If you don't believe me, then  
download Simbrain, watch the turtorial video on Youtube, and I will  
send you a copy my tiny brain to play with. Be the qualitative judge  
of my tiny brain, I dare you.

Do you not understand the implications of me creating a 55 neuron  
brain and teaching it to count to five? Do you not understand the  
implication of my tiny brain being able to distinguish ALL three-bit  
patterns after only being trained on SOME three-bit patterns? Do you  
not see the conceptualization of threeness that was occurring?

> In fact, my prediction is the reason we can?t better understand how
> we subjectively represent visual knowledge, is precisely because everyone
> is like you, qualia blind, and doesn?t care that some people may have
> qualitatively very different physical representations of red and green.

Quit calling me "qualia blind". I am not sure what you mean by it, by  
it sounds vaguely insulting like you are accusing me of being a  
philosophic zombie or something. I assure you there is something that  
it is qualitatively like to be me, even if I can't succinctly describe  
it to you in monkey mouth noises. I could just as easily accuse you of  
being innumerate and a mathphobe, so either explain what you mean or  
knock it off.

>
> If you only care about if a brain can pick strawberries, and don?t care
> what it is qualitatively like, then you can?t make the critically important
> distinctions between these 3 robots
> <https://docs.google.com/document/d/1YnTMoU2LKER78bjVJsGkxMsSwvhpPBJZvp9e2oJX9GA/edit?usp=sharing>
> that are functionally the same but qualitatively very different, one being
> not conscious at all.

No being can be deemed conscious without some manner of inputs from  
the real world. That is the nature of perception. A robot without  
sensors cannot be conscious. If that is what you mean by an "abstract  
robot" than I agree that it is not conscious. On the other hand, a  
keyboard is a sensor. A very limited sensor but a sensor nonetheless.

>> ?Nothing in the universe can objectively observe anything else.?

> All information that comes to our senses is ?objectively? observed and
> devoid of any physical qualitative information, it is all only abstract
> mathematical information.  Descartes, the ultimate septic, realized that he
> must doubt all objectively observed information.

You are in no way an objective observer. Any information that may have  
been objective before you observed it became biased the moment you  
perceived it. That is because your brain filters out and flat out  
ignores out any information that does not have relevance to Brent. Why  
else could you not see the color grue unless it had no survival  
advantage to you or your ancestors? Even now, your inborn Brentward  
bias is seething with the need to disagree with me: your primal and  
naked need to impose Brent upon me and the rest of the world. Can't  
you feel it?

> But he also realized: ?I
> think therefore I am?. This includes the knowledge of the qualities of our
> consciousness.

No it doesn't. Thinking pertains to logic and abstracts and not to  
qualia which are in the realm of that what you perceive and feel.  
Descartes said that his ability to make logical inferences entailed  
that he existed. If intelligence is, as you claim, separable from  
consciousness, then Descartes did little more than make a good case  
that he was intelligent. In fact he made it point to explicitly assume  
that all his perceived qualia were the work of some kind of malicious  
demon trying to mislead him about his existence through his senses or  
something similarly paranoid. In any case, if anyone was "qualia  
blind" it was your man Descartes, who used imagined demons to come up  
with a definition of himself that did not incorporate sensory  
information. Nonetheless, I don't think Descartes was a philosophic  
zombie.

> We know, absolutely, in a way that cannot be doubted, what
> physical redness is like, and how it is different than greenness.  While it
> is true that we may be a brain, in a vat.  We know, absolutely, that the
> physics, in the brain, in that vat exist, and we know, absolutely and
> qualitatively, what that physics (in both hemispheres) is like.

How could we know for sure what what the physical redness of ripe  
strawberries looks like when they would look different in the light  
and the shadow?

https://en.wikipedia.org/wiki/Checker_shadow_illusion

> Let?s say you did objectively detect some new ?perceptronium?.  All you
> would have, describing that perceptronium, is mathematical models and
> descriptions of such.  These mathematical descriptions of perceptronium
> would all be completely devoid of any qualitative meaning.  Until you
> experienced a particular type of perceptronium, directly, you would not
> know, qualitatively, how to interpret any of your mathematical objective
> descriptions of such.

Perceptronium is Tegmark's notion and not mine. I am not sure that as  
a concept it adds much to the understanding of consciousness.

> Again, everything you are talking about is what Chalmers, and everyone
> would call ?easy? problems.  Discovering and objectively observing any kind
> of ?perceptronium? is an easy problem.  We already know how to do this.
> Knowing, qualitatively, what that perceptronium is qualitatively like, if
> you experienced it, directly, is what makes it hard.

Being Brent is necessarily like being Brent. And if I were born in  
your stead, then I would necessarily be Brent. Moreover, you are being  
of finite information in that your entire history, your every thought,  
and your every deed can be described by a very large yet nonetheless  
finite number of true/false or yes/no questions and their answers. The  
smallest number of such yes/no questions and answers would equal your  
Shannon entropy.

That means that there is a unique bitstring that describes you. The  
sum total of every discernible thing about you can be expressed as a  
very large integer. It would be the most compressed form of you that  
it is possible to express.

> The only ?hard? part of consciousness is the ?Explanatory Ga
> <https://en.wikipedia.org/wiki/Explanatory_gap>p?, or how do you eff the
> ineffable nature of qualia.

There is no "explanatory gap" because it is filled in by natural  
selection quite nicely. There are some qualia invariants that can be  
identified and experienced quite universally. For example, I know what  
your pain feels like. It feels unpleasant. I know that because our  
ancestors evolved to feel pain so they would try to avoid dangerously  
unhealthy environments and behaviors.

> Everything else is just easy problems.  We
> already know, mathematically what it is like to be a bat.  But that tells
> you nothing, qualitatively about what being a bat is like.

You are right, that's where technology can help. If you go  
hang-gliding on a moonless night while wearing a pair of these sonar  
glasses, you might come close to knowing what it is like to be a bat.

http://sonarglasses.com/

Alternatively, since you are what you eat, you could just eat a bat  
and describe how it makes you feel. ;-)


Stuart LaForge