There is another meme “scientific” paper (well, it is a “research paper”, which does not have be correct lmao) about trying to interpret of what transformers actually do.
When the hype was at its peak, I wrote an article about “handwaving with too abstract math” or “sweeping the meaning under the rug”. I had very strong intuition that I have seen this before, and now I will show it.
Where all have seen this kind of sophisticated bullshitting with abstract entities taken out of context (from another, highly remote and ephemeral levels of abstraction) bing used to explain a natural phenomena? All the mythology and esoteric cults aside, provided they are “founded” on exactly this kind of “logic” (“how a loving and almighty God could ever let ya down?”).
We have seen it in Freudian so-called “psychology”, where concepts and even entities taken from Greek mythology (imagine!) were used to explain phenomena which has its causality at the level of melienated axons and releases of neuromodulators.
We have the same crap with employing abstract concepts of “Attention” and what not in the context of mapping words from one language to words of another /through corresponding aspects of the same reality, to which words are mere labels". At least this is what a valid translation ought to be.
Another place where we have seen “Transformations” of sophisticated abstract bullshit is, of course, Hegelian “philosophy” and his “logic”. Truly abstract concepts (without definite meaning) were use to argue about other abstract concepts, exactly as in Tantric religious cults.
Last but not least, there are “scientific” books and whole “areas of research” in which they are trying to interpret the Indo-Tibetan tantric mantras as if they are something other than a socially constructed and “randomly” selected by frequency of use in particular communities, strings of sounds.
This social phenomena - trying to interpret some socially constructed bullshit postfactum as having some deep and profound meaning is characteristic of pseudo-sciences and humanitarian “disciplines” in general.
Here is an example of a paper https://arxiv.org/pdf/2309.07315.pdf which tries to interpret what transformers do.
To do this lets first consider how they came to be. The answer, it seems, is that just as people tend to just throw some sophisticated terminology (they do not really understand) into their speech to look “competent” and “up to date”, people could throw some abstract mathematical operations (implications of which they do not really understand).
This is how the original paper has been produced. Look, we have thrown lots of abstract crap onto the wall and this is what stuck.
And now there is whole “are of research” of interpreting these findings and to show the profound meaning.
Yes, it is possible to parse everything (it is all math after all) int o less abstract (non-bullshit) universal concepts of “weights” and “distances”, just as it all started, before resorting to “handwaving” and to “what sticks”.
The use of the concepts of hyper-spheres (some abstract, non-existent notion of all the points at equal distance from the center) is the same Hegelian bullshit, but with mathematical notions. Let me show you why.
People tend to think with gross over-simplifications (which seem to “make sense”) because it is much more enrgy efficient and crude conceptual “idols” are good-enough.
One could readily imagine a surface of the sphere of the Big Bang explosion as (very naive) picture of reality and write some “research papers”. One could even throw the word “hyper” in it.
OK, lets talk about the shape (and many structures) of Reality.
It is not a arbitrary or random that we have observed so-calles “Tree Of Life” and that it has the form (shape) of a tree (of an acyclic graph – I am throwing some fancy terminology in). The Causality Itself leaves this kind of “tracks” which we could observe.
At all the edges of this “graph” is everything what is – all the individual molecular arrangements, including every single atom, molecule, dogs, cats, you and me.
All together we form a “front” of Existence which, one can readily imagine as some surface, but NOT a sphere by any means. Well, we cold say it is a sphere with respect to the abstract notion of time (we all at the same distance from so called Big Bang), but this abstract and irrelevant (orthogonal) to anything notion.
Ok, lets go back down to Earth. There are lots of structure in What Is. In fact - everything has a structure, and the Causality principle makes everything to look like a “Tree” (to an intelligent observer).
Now guess what? All the correct and valid translation has to be done by tracing a labeled by a sequence of sounds concept of the Mind from one language back to What Is (what it is a label of) and then relabel it in another language.
This is how human translators of the past used to translate everything, including mathematical and scientific texts. Every human translator, by the way, will tell you that one has to understand the context in order to translate correctly, and it is the context (the What Is) is that ultimately makes any translation possible (and valid).
So, there are abstract notions of relative distances from (and between the related concepts) and even of clusters around every single aspect of reality labeled by a human language, but there are no hyper-spheres, sorry, Chuds.
The abstract idea of “matching against the same structure of Reality” goes back to Upanishads and the Buddha and was the way to see things as they are and to call bullshit. Translation, of course, has to be based on the very same universal principle.
So, if we want to be less “Freudian” and “Hegelian” we have to consider the actual shapes and structures of reality, NOT what we could build using mathematics for “handwaving”.
One more time: there is no hyper-spheres anywhere in the Universe, so any interpretation which invokes them is bullshit.
And the only “distance” we have is a conceptual distance (of generalizations and resulting abstractions) from Reality itself.
And translation is simple - one unwinds the word back to What Is (including socially constructed crap) and then selects corresponding label from the target language. At least this is how all the Upanishads has been translated.
BTW, when one sees the distortion of the Buddha’s teaching by translators and interpreters into what we see and read today, one would see the model of the whole world of so-called “modern science” and AI “research”.