[ExI] Why stop at glutamate?

Tue Apr 11 08:53:15 UTC 2023

AI color training. Different hierarchies (not qualia???), patterns
represent the colors, and so on....

https://openaccess.thecvf.com/content_CVPRW_2020/papers/w26/Hickey_Hierarchical_Color_Learning_in_Convolutional_Neural_Networks_CVPRW_2020_paper.pdf

On Tue, Apr 11, 2023 at 1:50 AM Giovanni Santostasi <gsantostasi at gmail.com>
wrote:

> See, without even knowing this field at all but having worked in
> neuroscience I correctly predicted how a neuroscientist would talk about
> this business of colors. It is all about how the neural patterns happen, in
> which sequence, which brain regions are involved and so and on. Not one
> single word as "quality" appears anywhere.
>
> [image: image.png]
>
> On Tue, Apr 11, 2023 at 1:45 AM Giovanni Santostasi <gsantostasi at gmail.com>
> wrote:
>
>> The visual system encodes color by means of a distributed representation:
>> the activity of many neurons preferring different colors, but with broad
>> and overlapping tuning curves. This means that similar colors evoke similar
>> patterns of activity, and neural representations of color can be
>> characterized by low-dimensional “neural color spaces” in which the
>> positions of colors capture similarities between corresponding patterns of
>> activity (Brouwer and Heeger, 2009
>> <https://www.jneurosci.org/content/33/39/15454#ref-5>).
>> Categorical Clustering of the Neural Representation of Color
>> https://www.jneurosci.org/content/33/39/15454
>>
>> On Tue, Apr 11, 2023 at 1:37 AM Giovanni Santostasi <
>> gsantostasi at gmail.com> wrote:
>>
>>> Another important paper. The AI actually came up with the idea of color
>>> in an emergent way. Patterns is all what there is.
>>> Emergent color categorization in a neural network trained for object
>>> recognitionhttps://elifesciences.org/articles/76472
>>>
>>> On Tue, Apr 11, 2023 at 1:22 AM Giovanni Santostasi <
>>> gsantostasi at gmail.com> wrote:
>>>
>>>> Ok here is a paper that should clarify a lot on this topic of color
>>>> perception.
>>>> Let's read it and see what we get from it. It seems to me that again
>>>> they simply used fMRI (so basically brain pattern) to deduce all what there
>>>> is to deduce about redness and the similar.
>>>> Neural representations of perceptual color experience in the human
>>>> ventral visual pathwayhttps://www.pnas.org/doi/10.1073/pnas.1911041117
>>>>
>>>>
>>>> On Mon, Apr 10, 2023 at 4:48 PM Jason Resch via extropy-chat <
>>>> extropy-chat at lists.extropy.org> wrote:
>>>>
>>>>>
>>>>>
>>>>> On Mon, Apr 10, 2023, 7:08 PM Brent Allsop via extropy-chat <
>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>> On Mon, Apr 10, 2023 at 11:11 AM Jason Resch via extropy-chat <
>>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>>
>>>>>>> On Sun, Apr 9, 2023 at 5:20 PM Giovanni Santostasi via extropy-chat <
>>>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>>>
>>>>>>>> If this doesn't destroy completely anybody illusion that the a
>>>>>>>> brain made of meat (and particular stuff like glutamate) I don't know what
>>>>>>>> else it could. These people will always believe that meat brains are
>>>>>>>> necessary because God made them so. No amound of science would convince
>>>>>>>> them.
>>>>>>>>
>>>>>>> 2) You can train an AI to recognize activation patterns in the brain
>>>>>>>> and associate them with particular stimuli. This has been tried with words
>>>>>>>> and even images both in wake and dreaming state. Here an example that
>>>>>>>> should blow everybody minds:
>>>>>>>> https://www.biorxiv.org/content/10.1101/2022.11.18.517004v2.full.pdf
>>>>>>>> Again, from this study we can see that it doesn't matter how the
>>>>>>>> pattern is generated, but that there is a pattern of activation. These
>>>>>>>> patterns are unique for each individual but statistically they are similar
>>>>>>>> enough that after training over many subjects you can give a statistical
>>>>>>>> estimate that the person is seeing or even thinking about something in
>>>>>>>> particular. Again, IT WORKS people !
>>>>>>>>
>>>>>>>
>>>>>>> I consider this a knock-down argument against the functional role of
>>>>>>> glutamate (or other molecules) in the sensation of red. These tests use
>>>>>>> only blood flow data, which is a proxy for neural activity. They are not
>>>>>>> measuring ratios of specific neurotransmitters or molecules, or
>>>>>>> introspecting the activity within the cell, the fMRI looks only at which
>>>>>>> neurons are more vs. less active. And yet, from this data we can extract
>>>>>>> images and colors. This proves that neural activity embodies this
>>>>>>> information.
>>>>>>>
>>>>>>
>>>>>> I guess I've failed to communicate something important about why we
>>>>>> use glutamate.  The primary reason we use glutamate is precisely because of
>>>>>> its ease of falsifiability.  I fully expect redness to be falsified
>>>>>> (someone will experience redness with no glutamate present) and something
>>>>>> different from glutamate will then be tried, and eventually something will
>>>>>> be found to be experimentally proven to be redness.  Easy and obvious
>>>>>> falsifiability is what everyone is missing, so THAT is what I'm most
>>>>>> attempting to communicate with the glutamate example.
>>>>>>
>>>>>> If you guys think there are knock down arguments for why a redness
>>>>>> quality is simply due to recursive network configurations (I am not yet
>>>>>> convinced, and am still predicting otherwise (see below), and it's much
>>>>>> easier to say glutamate than whatever stuff you guys are talking about,
>>>>>> which nobody is concisely stating, and I have problems understanding), then
>>>>>> please, every time I say 'glutamate', do a substitution for anything you
>>>>>> like such as  'Recursive network model A', or any other yet to be falsified
>>>>>> theory.  And let's leave it up to the experimentalists to prove who is
>>>>>> right, like good, humble, theoretical scientists should.
>>>>>>
>>>>>>
>>>>>> P.S.
>>>>>> At least that paper
>>>>>> <https://www.biorxiv.org/content/10.1101/2022.11.18.517004v2.full.pdf> you
>>>>>> referenced has pictures (composed of real qualities), not just abstract
>>>>>> text (tells you nothing about qualities), as text only would be completely
>>>>>> meaningless, right?
>>>>>> But why don't you guys ask the publishers of that paper, how they
>>>>>> came up with the qualities displayed on the images depicting what they are
>>>>>> detecting?
>>>>>> Here is a link to Jack Galant's work
>>>>>> <https://www.youtube.com/watch?v=6FsH7RK1S2E&t=1s>, done over a
>>>>>> decade ago, to which all these modern examples are just derivative works,
>>>>>> easily done with modern AI tools.
>>>>>> When I saw Jack Galant's work
>>>>>> <https://www.youtube.com/watch?v=6FsH7RK1S2E&t=1s> back then, I knew
>>>>>> he had a problem determining what qualities to display on his screens,
>>>>>> depicting what he was detecting.  The fMRI only providing abstract
>>>>>> qualityless data which is meaningless without a quality grounded dictionary.
>>>>>> So I called him and asked him how he knew what qualities to display.
>>>>>> He immediately admitted they "false-colored" them (Jack Gallant's words).
>>>>>> They used the original color codes in the digital images they were showing
>>>>>> to their subjects, to determine what color to display.  In other words,
>>>>>> they were grounding their colors to physical light, which is nothing like
>>>>>> either the properties of a strawberry, which the light merely represents,
>>>>>> or the very different properties of conscious knowledge they are detecting
>>>>>> and describing with qualityless abstract text.  As Giovanni admits, they
>>>>>> are correcting for any changes in physical properties or qualities they are
>>>>>> detecting so they can falsely map all those diverse sets of properties they
>>>>>> are detecting back to the same false colored light, blinding them to any
>>>>>> possible inverted qualities they may be detecting in all that diversity.
>>>>>>
>>>>>> By the way, I added this Japanese paper
>>>>>> <https://www.biorxiv.org/content/10.1101/2022.11.18.517004v2.full.pdf> to
>>>>>> the list of yet another example of quality blind papers, including Jack
>>>>>> Galant's work that only uses one falsely grounded abstract word for all
>>>>>> things representing 'red' here
>>>>>> <https://canonizer.com/topic/603-Color-Exprnc-Observation-Issue/1-Agreement>
>>>>>> .
>>>>>>
>>>>>> If anyone finds a peer reviewed paper that is not quality blind.
>>>>>> (other than mine
>>>>>> <https://www.dropbox.com/s/k9x4uh83yex4ecw/Physicists%20Don%27t%20Understand%20Color.docx?dl=0>,
>>>>>> which is about to be published) will you please let me know about one?  As
>>>>>> I will trust someone that believes and understands that qualities are
>>>>>> necessarily real properties of real hallucinations in our brain.  I predict
>>>>>> they are just the physical properties they are detecting but only
>>>>>> abstractly describing and then false coloring.
>>>>>>
>>>>>
>>>>>
>>>>> Brent,
>>>>>
>>>>> I appreciate that added detail and correction. If the colors in the
>>>>> reconstructed images are false colors or inferred by the AI from the
>>>>> reconstructed image then I retract my statement of it being a knockdown
>>>>> argument against the molecular basis of color qualia. I still suspect color
>>>>> information is encoded in the patterns of neural activity, but it may be at
>>>>> a low enough level that the fMRI lacks the spatial resolution to detect it.
>>>>>
>>>>> Jason
>>>>> _______________________________________________
>>>>> extropy-chat mailing list
>>>>> extropy-chat at lists.extropy.org
>>>>> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230411/c787338a/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 42100 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230411/c787338a/attachment.png>