Holographic theory of learning | Hacker News Reader

Holographic theory of learning

jovex.substack.com

59 pointsTomte37 comments

It's a nice read but I doubt that the idea delivers what it promises. I think one could downsize the idea to: if you learn one topic deeply you (probably) touch surrounding topics. That does not sound as elegant and holistic as in the article but is IMO closer to the truth. I would call it a huge exaggeration to say I've learned things about the world when in reality I just became a really good software engineer that knows how to interact with OSS communities.

These ideas are repeated often and I lean more to the specificity side of things: you only get good at what you learn/train. You won't become better at decision maker by learning chess/poker, you won't (or just become a slightly better) endurance swimmer by becoming a good runner, you won't understand human psychology by getting good at coding.

I remember a talk of top-notch mathematics where they were asked about related mathematical topics and most of them would just answer something like: "I just try to understand my field of mathematics well, I can't say much about something else". This was the discussion from the Breakthrough Prize in Mathematics 2015: https://www.youtube.com/watch?v=eNgUQlpc1m0&list=PLyF3OMOiy3...

This is the correct answer. The rest of the replies are cope from people who do nothing but study CS and SWE and think that therefore makes them knowledgeable of everything.

This theory was proposed by neuroscientist Karl Pribram in "Languages of the Brain": https://archive.org/details/B-001-014-181, probably 1971

Nice thoughts. I think the fictional writer example is the clearest example, as a writer creates worlds within the real world.

Maybe this is could be more about information compression? The author notes that the sliver of the "hologram" doesn't necessarily contain the whole, but it's an interesting idea. If you choose a topic, there are minor aspects of much larger ideas. You then have to decompress as you learn more about the context of the information. Like in the fictional writer example, you move outward from the focus and learn more about say the history of when the writing was completed. There will be elements of that context in the writing, but clearly not as much, because the information density of the fictional work is clearly smaller and cannot contain the entire context.

Now I wonder how holograms and compression are related...

Patterns, data repetition, self similarity. The universe loves to do this. I.e, Golden ratio. Light waves, ocean waves, snowflakes, etc.

Indeed, the fractal geometry of nature.

The world is fractal. The closer in you zoom, the more detail you see, possibly to an infinite level. So, yes, you can certainly learn a lot about everything, if you learn everything about something.

The world isn’t fractal. God made the universe with specific laws that operate on infinitesimal things many times a second. Mindboggling amount of objects, like subatomic particles or energy, that all operate at once across the universe. You zooming in and out just controls what percentage you see. It reflects our limits, not the universe’s.

It's a bit presumptuous to state that the world is or isn't anything.

It’s just a statement of belief. If we didn’t state beliefs, we wouldn’t be discussing anything of substance on Hacker News.

Whether it’s right or wrong is a different matter.

> And once you’ve acquired enough background information, you’ll learn even more through direct studying of his works. Each work you study, each novel for example, contains an entire fictional Universe with its logic and laws, and the characters of the novel are subject to all of them. Each such fictional universe is in some way an imperfect reflection of our own real universe, and so learning about how the fictional world functions in a novel, will tell us quite a bit about our real world.

Which is of course what LLMs depend on.

You might even call this moment in AI the Twingularity.

There is an entire science yet to be formalized in this essay. When the author begins to backtrack on their thesis, and then says:

> I don’t actually believe this to be a universally applicable principle, as there are lots of exceptions, but I feel that there is “something” about it that deserves our attention.

That "something" is identifying where our current sciences, our language, and our intellectual curiosities have not formalized investigation. The mere fact that this observation has no name that can be commonly referenced is the clearest indicator it's an overlooked aspect of living a life.

I wonder if these ideas are not formalized already, some philosophical school of thought labeling this with some too many syllable name. I also wonder if the Dunning–Kruger crowd's work touches on this aspect.

One formalising "something" could be category theory, in which we say we don't care too much about what objects are like qua objects (indeed, we'll handwave many things away "up to isomorphism"), but we do care greatly about the structure preserving maps ("conceits" in the lit crit language) we can find from other objects to a given object, or conversely from a given object to other objects.

(if we're really hardcore, we'll even claim the objects don't even have much left to study after we've learned all that we can about the identity maps from each object back to itself)

Exercise for the reader: what is the structural equivalent of an epimorphism in Snow's other culture? a monomorphism?

(quick sanity check: do the humanities even have a way to talk about domains and codomains? Yes: "signifier" and "signified". What about cancellation?)

Lagniappe: https://www.youtube.com/watch?v=iZzVdomYHnI

N.b., an Aristotelian substance is such that the form of the integral whole is in every part so to speak. This distinguishes integral wholes or substances from accidental wholes, like machines where the parts are just so arranged so that they happen to enter into certain causal relations. But the formal cause is strictly accidental; in true substances, in integral wholes, a substantial form renders the thing a true unity. The causal relations don't just-so-happen to be there, they don't just happen to result in an effect by virtue of their circumstance; they are intrinsically ordered toward an end.

And, of course, when you learn something, universal principles are presupposed by anything we study. Metaphysics, for example, studies the first principles of being qua being, and so whatever is known in a thing that is universally true or true of the class, is true of everything else, or of the class.

Furthermore, much of our knowledge is analogical. Thus, when we learn a new thing, we may do so by analogy or we may notice analogies. And let us not forget the Logos in which (or Whom) all finds its analogical being. The effect resembles in some sense its cause, and all that exists stands in an analogical relation to its creating and sustaining cause, the Logos.

I think that there's a lot of work in this area. In some sense, complexity theory itself is this approach applied to dynamic systems. You cannot fully understand the system by studying the individual parts, so if you want to understand something deeply, you need to study it holistically. Because everything in reality is necessarily intertwined with everything else, you keep hitting adjacent fields.

If there’s holistic detectives and assassins, this does indeed suggest the existence of holistic teachers and students :)

This seems highly speculative to me. Is there any research that suggests what is being described is true?

I'd expect research by lumpers would tend to find it true; research by splitters, false.

Sort of. Researchers were sent in to study it but with no upper boundary and so they kept at it. They went so deep in their research they ended up with the truth of the universe.

I've thought along these lines myself and wished that we could make more use of it because many times you only realize the pattern you just solved was actually a variation of some other pattern from another field after the fact.

The more I learn about one thing, the more my brain connects other things together

As the old saying goes, “all roads lead to Rome.”

But then what is Rome? I claim that it is God, the Summum Bonum, the source and the destination of all that is. But not merely as a culmination.

Human knowing involves a great deal of analogy, and nowhere else is analogy more essential than knowledge of the Highest Principle and First and Greatest Cause; here the so-called analogia entis. Thus, through knowledge of what is created, we come to know, analogically, the Creator, the Logos. As the metaphysical principle goes, the effect resembles in some way its cause, and this resemblance is analogical. Otherwise, we would need to choose between univocity and equivocity: the first leads the pantheism, the second to deism. In the first, we are God. In the second, God has nothing to do with us. But in the analogia entis, God stands in an analogical relation with us.

Fantastic article, thank you.

Wow. Great ideas.

Yeah go ahead and explain how learning Mandarin teaches you about Quantum Mechanics or how learning Music Theory helps you bake cakes. Cute idea that falls apart under casual inspection.

Music Theory and cake baking are both all about strict proportion. (and therefore the pythagorean comma would be the equivalent to how few sigfigs we need to keep when converting recipes in freedom units to g and kg?)

You are confused, you are learning arithmetic to help you understand recipes. Knowing that water plus yeast plus flour in a certain proportion = bread does not teach you how to compose music and vice versa.

Explain how to learn Mandarin in a way that avoids being exposed to, and learning from, texts that indirectly teach about QM.

I'm sorry but have you ever learned a language? It what universe do you need to read a text about QM to learn spanish??

Music theory deals with the harmonic series, which studies the motion of sound waves [1]. Sound waves are physical, and are affected by the matter and density of the particles they are traveling in. To understand it deeply, you need to get into the behavior of solids and gases, which is chemistry, something useful to baking.

Similarly from Mandarin to Quantum Mechanics you can take Mandarin -> linguistics -> Probability Theory[2] -> Quantum Probability -> Quantum Mechanics

1. https://en.wikipedia.org/wiki/Harmonic_series_(music) 2. https://www.pnas.org/doi/full/10.1073/pnas.1100760108

More directly, here's superposition (as opposed to classical pure states) in hanzi: https://languagelog.ldc.upenn.edu/nll/?p=64175

[If Google's xlation can be believed, note that syntactically the 3rd and 5th columns repeat in:

  吾生也有涯
  而知也无涯

yet semantically they group differently between the lines:

  吾生/也有/涯
  而/知也/无涯

compare resonance in aromatics]

This strains credulity. You do not by osmosis learn how to solve the wave equations or manipulate them by playing the Tuba.

What you are doing is connecting related fields however tangential they may be and implying that learning about one makes you learn about the others which is just not true. Let's take a Music Theory student and give them a series of problems from Landau and Lifshitz, how well do you think they will do?

I think that i'm illustrating the point of the author. That every field is connected with every other field. If you start getting into the harmonic series as a music theory nerd, you will be aided by studying physics and chemistry. Maybe your average music student wouldn't but John Coltrane certainly did draw these types of connections.

" Everything is deeply intertwingled. In an important sense there are no "subjects" at all; there is only all knowledge, since the cross-connections among the myriad topics of this world simply cannot be divided up neatly."

-- Theodor Holm Nelson

( https://en.wikipedia.org/wiki/Intertwingularity )

  First come I; my name is J—w—tt:
  There's no knowledge, but I know it.
  I am the Master of this college;
  What I don't know isn't knowledge.

ca. 1880: https://en.wikipedia.org/wiki/Balliol_rhyme

Reply on news.ycombinator.com