<div dir="ltr"><div>Hi Stuart,</div><div>Yes, thanks for this reference.</div><div>It just must be kept in mind, that though an abstract intelligence (words only) can discover and model color qualities, without a definition of the words being grounded with factual physical qualities it experiences first hand, they can't know what the words represent.</div><div><br></div><div><img src="cid:ii_mby04kis0" alt="The-Strawberry-is-Red-0480-0310.jpg" width="480" height="310" class="gmail-CToWUd gmail-a6T" tabindex="0" style="cursor: pointer; outline: 0px;"></div></div><br><div class="gmail_quote gmail_quote_container"><div dir="ltr" class="gmail_attr">On Sun, Jun 15, 2025 at 11:16 AM Stuart LaForge <<a href="mailto:avant@sollegro.com">avant@sollegro.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Here is an interesting study that reports using machine learning <br>
techniques to analyze color qualia quantitatively. Basically they had <br>
426 people with typical color vision and 257 individuals who were color <br>
blind take a computer survey where they judged the similarity of <br>
color-pairs chosen randomly from a pool of 93 colors using a point <br>
scale. The results were then used to train an artificial neural network <br>
(ANN) by unsupervised pairwise alignment of individual's similarity data <br>
for the color pairs without reference to the color name or label. In <br>
other word, it was told to look for similarities based on the numerical <br>
distance reported by each pair of individuals for each color pair, <br>
without being told what the colors were and then used the alignments to <br>
form clusters corresponding to a "color map". When the relative <br>
differences between colors reported by the research subjects were <br>
clustered without reference to the color, it nonetheless turned out that <br>
clusters corresponded to the various color labels and the color maps of <br>
the normally-sighted group were similar to one another. The color maps <br>
of the color-blind people were, also, similar to one another. However, <br>
the color maps of the color-sighted people were different from the color <br>
maps of the color-blind people.<br>
<br>
<a href="https://www.cell.com/iscience/fulltext/S2589-0042(25)00289-5" rel="noreferrer" target="_blank">https://www.cell.com/iscience/fulltext/S2589-0042(25)00289-5</a><br>
<br>
<a href="https://www.lesswrong.com/posts/LYgJrBf6awsqFRCt3/is-red-for-gpt-4-the-same-as-red-for-you" rel="noreferrer" target="_blank">https://www.lesswrong.com/posts/LYgJrBf6awsqFRCt3/is-red-for-gpt-4-the-same-as-red-for-you</a><br>
<br>
Using machine learning to analyze qualia like this is fascinating. Just <br>
like an LLM can learn the contextual meaning of words without being <br>
explicitly programmed with the definition of the words simply by <br>
statistically analyzing the average numerical distances between words in <br>
a corpus of text, this technique should allow AI to recognize and use <br>
colors without being explicitly programmed with any particular <br>
definition of say red. This would render the question of whether an AI <br>
can truly see a color to be equivalent to whether an LLM actually <br>
understands what it is saying.<br>
<br>
Brent, you have have a thing for both color qualia and surveys so this <br>
paper should be right up your alley.<br>
<br>
Stuart LaForge<br>
<br>
-- <br>
You received this message because you are subscribed to the Google Groups "extropolis" group.<br>
To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:extropolis%2Bunsubscribe@googlegroups.com" target="_blank">extropolis+unsubscribe@googlegroups.com</a>.<br>
To view this discussion visit <a href="https://groups.google.com/d/msgid/extropolis/fe0026617a15416bc12866444518c111%40sollegro.com" rel="noreferrer" target="_blank">https://groups.google.com/d/msgid/extropolis/fe0026617a15416bc12866444518c111%40sollegro.com</a>.<br>
</blockquote></div>