Hacker News Reader

Comment by marginalia_nu | Hacker News Reader

Yeah. There are serious epistemological problems to go with externalized knowledge.

Knowledge isn't just the ability to produce true statements, it isn't knowing what things are or what things will be, it's knowing why things are and why things will be.

If I keep guessing coinflips, I don't have knowledge of the answer 50% of the time even though I can predict them that often.

Knowledge has an aspect of understanding why things are, not just that they are, and that aspect becomes incredibly weak with the "look up facts on wikipedia"-model of external knowledge.

The JTB-model of knowledge is probably incomplete, but it's arguably less wrong than a model that doesn't contain justification.

View on HN

> If I keep guessing coinflips, I don't have knowledge of the answer 50% of the time even though I can predict them that often.

Sorry to get off topic, but I think this is an interesting way of thinking about this. I wonder if it represents any kind of larger difference in our worldviews. I would say that we have knowledge of the answer 0% of the time. If we guess and it happens to be correct, that's a coincidence, not an indication of knowledge.

IMO this goes hand in hand with thoughts I've had about the concept of mistakes. When I used to dabble in day trading (I've been clean for a few years now), I gained and lost and gained a lot of money. Whether any particular gamble ended up as a gain or a loss, I consider them all mistakes because I had no rational reason to expect any of them to pay off.

My uneducated hypothesis is that whether somebody thinks of a successful gamble as a mistake or a good decision could be a decent predictor of certain personality traits and political views. Maybe the same applies to the question of whether being correct implies knowledge.

The JTB model is wrong or at least misleading. Definitely outdated.

Our best modern models for both biological and artificial intelligence indicate that concepts or knowledge* emerges from networks of sensation/facts/data.

* Specifically referring to "system 1" cognition, which is what the article is ostensibly targeting.

Well the JTB was the standing model for some 2300 years, so even though there are a few corner cases it doesn't quite cover (Gettier etc), those are corner cases indeed. It doesn't invalidate JTB any more than Einstein invalidated Newtonian physics.

The question is what we mean when we say knowledge. I don't think modelling is a good way of answering that question. Of course there is going to be a connection between perception and knowledge, how else would the knowledge enter us? But David Hume could have told you that.

Knowledge doesn't "enter" us, that's the point. It's an emergent property of (many, many) memorized & networked sensations, which do enter us through our various sensory organs.

It is virtually impossible to teach an abstract concept like "cup" without a) providing real, sensory examples of what you mean by "cup" or b) relating to analogous concepts that the student has already learned through personal experience (like "bowl" with "handle") and is capable of communicating.

There is a point where we do not have knowledge, let's say when we are born, then we have perceptions, and after that we (may) have knowledge about the world. I will say that knowledge has entered us through perceptions. If something is not in my mind, I have a perception, then it is in my mind; and if I was unable to have perceptions it could not have entered my mind, then it has entered through that perception. All physical phenomena are to some extent emergent down to subatomic particles and possibly even further. It doesn't really have much of an effect on the subjective human experience which principle is more emergent, and the subjective human experience is the only human experience.

It's very hard to teach someone who has no relationship with the world, but you don't need a large set of concepts to synthesize additional concepts. You can derive most of mathematics from a few simple axioms. Democritus concluded the world was made of atoms based on observations about how the world appeared to work, on encountering problems like Zeno's paradox. Knowledge, as well as language, is all about how things relate to each other. Words to meanings, causes to effects.

> You can derive most of mathematics from a few simple axioms.

Yeah, nothing is stopping you from building towers of abstractions that are N degrees from any real experience or data. But those heavily-derived concepts are prohibitively difficult to teach and communicate, because each level in the abstraction hierarchy adds semantic noise. A major reason why classical, "pure" mathematics pedagogy is infamously ineffective.

Ultimately, every great mathematician learned to count with rocks or apples -- not by internalizing Peano's axioms.

That is how it is commonly taught, but can it really be argued it's the only way it can be taught? Regardless which method is easier, why is it possible to interact with negative, irrational, or complex numbers without anyone having ever seen them, but not integers?

There of course needs to be a common set of ideas to communicate, but I'm not convinced you need one particular set. In many cases, having some of the ideas means you can synthesize the rest. You could for example reason about numbers by drawing upon size rather than quantity. Then you basically end up with the Euclidean method, a mathematics of proportion that goes a surprisingly long way.

Abstract quantities, even positive integers, don't exist in reality. Heck, even "objects" don't really exist, because "boundaries between things" don't concretely exist except in our perception and imagination.

If there is any primal, axiomatic concept, it is "object"/"thing". From object you can derive quantity (many objects), from quantity you can derive quality (two objects are different), and so on and so forth.

You can reason about the same reality in various different terms, and the choice in terms shapes how you see the world. I don't think it's easy to argue that one concept is the true concept from which all concepts inevitably derive. You can just as easily think of things as parts of a whole as objects in an emptiness.

In some languages the name for door is the same as the name for mouth, and a door is a bit like a mouth so it makes sense to relate them. All language is metaphorical like that, except not always as explicitly. When we say a thing is an object, we say that our concept of that thing shares similarities to the concept we have for objects.

> It's very hard to teach someone who has no relationship with the world

It is impossible, by definition.

Reply on news.ycombinator.com