[FRIAM] serious AI question

Prof David West profwest at fastmail.fm
Mon Jan 30 16:02:56 EST 2023

This issue needs some further background. I arose from an ongoing discussion with Jenny Quillien (FRIAM) and Richard Gabriel (not of FRIAM) with regard aesthetics.

For Jenny, the immediate concern is with a subset of Zen Art (painting, poetry, ceramics, calligraphy) that evokes," triggers, engenders an "Aha!" state of mind, or instills a sense of "Oneness." Alternatively, it might be said that one encountering such a piece of art can say with certitude, "yes, that creator gets it!"

A story to illustrate: When the Fifth Patriarch of Zen knew his death was imminent, he bade his acolytes to express their understanding of Zen in the form of a poem. Only the lead student did so, and he did it anonymously by writing it on a wall at night. Hui Neng had some one read the poem to him, and recognized that the author did not understand Zen at all and had another person write a poem he dictated on the opposite wall. The Fifth Patriarch read the poems on both wall and knew that Hui Neng was to be his successor.

A longstanding conversation with Richard (and Jenny) revolves around Christopher Alexander's notion of QWAN—Quality Without A Name. In later work, Alexander used the term "liveness" as a synonym/successor to QWAN. In both cases is is a quality or an attribute whose presence or absence makes a building (any human built environment) "beautiful" or not.

Alexander asserts that QWAN (liveness) is universal and timeless. He did a lot of experiments with oriental carpets to test his assertion. Richard repeated Alexander's experiments and did other with photographs and artifacts. All of these experiments, statistically, seemed to prove Alexander correct. Statistically, because there are always curmudgeons who thing the Taj Mahal is just a tomb.

An AI came into the discussion: 1) could you build/train an AI (ZenChatGPT) to write a poem, ala Hui Neng, that appeared to embody a "true understanding of Zen?; 2) could you build/train an AI to sort thru the Google image base and detect "Zen evocative art" or buildings with QWAN/liveness?


The two poems— which one was Hui Neng's?

        The body is the wisdom-tree,
        The mind is a bright mirror in a stand;
        Take care to wipe it all the time,
        And allow no dust to cling.

        Fundamentally no wisdom-tree exists,
        Nor the stand of a mirror bright.
        Since all is empty from the beginning,
        Where can the dust alight

On Mon, Jan 30, 2023, at 11:46 AM, glen wrote:
> While I appreciate Jochen's and SteveS' responses, they didn't cover a 
> point I think might be useful. You point out the assumption of 
> effability, which is interesting. But I think a more fragile assumption 
> is that there *are* categories in humans/artifacts that are 
> classifiable in the first place. I think we can safely leave aside that 
> humans and the artifacts they respond to are different. But if others 
> feel it makes a difference, I'm happy to lob some words at it.
> On one extreme of a spectrum, let's say the left, we could place the 
> situation that there is 1 type of human/artifact that trigger in this 
> way (implying there are 2 types, those who do and those who don't). On 
> the other end of the spectrum, let's say the right, there are as many 
> classes/categories as there are humans/artifacts. I.e. any possible 
> human/artifact might be triggered, depending on the circumstances. But 
> no 2 people will trigger under the same circumstances.
> My claim is that AI/ML will *not* be useful in an open set from the 
> middle to the right. The lower bound will move as we apply more 
> complicated AI/ML. But at the limit, if everyone's in their own class, 
> they're not really classes. But it may be useful for a region on the 
> left.
> I think DaveW's also assuming that there *are* classes to find. Whether 
> I buy that assumption or not is irrelevant. I think it needs to be 
> defended. Why would we believe there are classes of human/art that 
> would trigger this, rather than, say, a random event where the 
> humans/artifacts are rationalized later as having been triggered and 
> been the trigger?
> On 1/28/23 15:10, Prof David West wrote:
>> This is a serious question albeit one in a realm that many would dismiss as non-serious. First, some background.
>> Rinzai Zen is the "sudden enlightenment" school that asserts the possibility of a single event serving as a 'trigger' that evokes/instills-in-the-mind a state of enlightenment. The trigger might be a closed fist of your guru striking your ear, or—as was the case with Hui Neng (illiterate peasant who became the Sixth Patriarch) overhearing a fragment of the Diamond Sutra spoken by a passerby of the fish market where he was working.
>> This kind of "evocative trigger" is analogous to your nose detecting the scent of cinnamon as you walk past a bakery and your mind instantly filled with a complete memory of grandmother's kitchen, all the scents and sounds, and emotions, an activities, in complete detail.
>> A 'Zen evocative trigger' would, by analogy, fill your mind with—put your mind in a state of—Enlightenment. This might be ephemeral, satori with a lower case 's', or permanent, Satori with an upper case 'S'.
>> There is a large body of art (calligraphy, painting, poetry, ceramics, ...) that embodies exactly this kind of trigger; one that can be 'sensed' even if its sensing does not trigger (S)satori.
>> So the question: is it possible to construct a self-learning AI with a training set of such art and, once trained, turn it loose on the Google image base to find other examples of art with evocative triggers?
>> Of course, there is a hidden assertion: whatever the quality or characteristic of the art that embodies the 'trigger' is ineffable; which means, in this case, it has no "representation" (word, symbol, brush stroke, etc.).
> -- 
> ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ
> -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom 
> https://bit.ly/virtualfriam
> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIC http://friam-comic.blogspot.com/
> archives:  5/2017 thru present 
> https://redfish.com/pipermail/friam_redfish.com/
>   1/2003 thru 6/2021  http://friam.383.s1.nabble.com/

More information about the Friam mailing list