Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Always Bet on Text (2014)

graydon2.dreamwidth.org

Post reply on HN
Print view View original

331–340 of 349 posts

Re: Always Bet on Text (2014)

#331

Post by johnfn »

Nitramp wrote: post #304
aftbit wrote: post #268

As usual, context matters. If you're trying to convey complex thoughts about abstract matters, then language is the way to do it. If you are trying to convey deep emotional states, then a photo or video is probably better. You cannot convert the 1st amendment into photos, and you cannot convert the Mona Lisa smile into words. Just try to use words to describe either of these photos and see how they fall short: https:…

John Steinbeck arguably did that for your first image: https://en.m.wikipedia.org/wiki/The_Grapes_of_Wrath It's not really apples to apples of course. Steinbeck describes much more, but also takes much more time to consume. The point though is that it's possible to convey very complex emotional states in literature or poetry

I mean, if your best argument is that one of the greatest authors of a generation needed an entire book to capture the emotion of a single picture, that argument doesn’t seem very strong to me :)

Re: Always Bet on Text (2014)

#332

Post by DarkWiiPlayer »

duckerude wrote: post #214
DarkWiiPlayer wrote: post #205

Earlier quoted context omitted.

SVG is XML, which I'd say counts as text.

I don't think it counts for the purposes of this blog post. You can't translate the Wikipedia quote in the article into an SVG except by just rendering the bare text. Checking programmatically that text is about "bird" or "Twitter" is very easy. Checking that an SVG is about "bird" or "Twitter" is very hard. It's not even efficient. It conveys the same message as the word "Twitter" in a hundred times the size. And it…

> Checking programmatically that text is about "bird" or "Twitter" is very easy. Checking that an SVG is about "bird" or "Twitter" is very hard.

That depends on what you want:

If you know what twitter is, then yes, just saying "twitter" is more efficient than a whole SVG file describing the twitter logo.

But if you don't know what twitter is, nor what its logo looks like, SVG is a very efficient way to describe the logo exactly. If you can have it rendered to your screen that's optimal, but if not, you can take it as instructions to redraw the logo yourself and still get an accurate impression of what the twitter logo looks like.

I that sense, "twitter" and the SVG of the twitter logo are both text, but they don't have the same meaning.

As for the granite, it is indeed one situation where just drawing the plain SVG logo is more effective, but SVG would still be viable. Over telephone, on the other hand, you just can't transmit an image. You can only attempt to describe it, and the more detailed you get with that, the closer you come to just using a code very similar to SVG.

Re: Always Bet on Text (2014)

#333

Post by necovek »

KMag wrote: post #25
necovek wrote: post #19

Earlier quoted context omitted.

That's really interesting, so thanks! However, I am not sure if that is supposed to affect my opinion of his take on text vs other media? (I try not to let prejudices like "he's a really smart person" affect me when discussing a topic that is very approachable to a wider population)

It hopefully doesn't affect your opinion on his take, but it hopefully gives some context related to his background. He's got a bunch of experience in very low-level very back-end systems, involving a lot of abstract reasoning. I would guess people used to a lot of abstract reasoning would tend to be partial to plain text and/or mathematical notation.

Cool, that's a very interesting observation, so thanks for elaborating!

As originally worded, it only sounded like "hey, this dude is famous" (for HN value of famous :-), so I would appreciate the commentary you just shared right off the bat! Thanks again for the clarification!

Re: Always Bet on Text (2014)

#334

Post by eimrine »

soapdog wrote: post #196

It really depends on what you define as text. Does it involve special notations such as math? Can it involve diagrams? If all you mean is prose, as writing sentences in whatever native language you speak, then I must respectfully disagree with the OP. Yes, text is marvelous. I'm an author of multiple books and have a passion for the written word, but there are things for which there are better ways to convey informat…

Let me try. Text is anything human-readable which supports easy editing in WYSIWYG way. For example: XML, maybe JSON, non-obfuscated source code, LaTeX notation for math, vector images, chess notation, etc. Diagrams are not text but they are easy to be describen in plain English, much easier than memes or photos.

Electric circuit is a beautiful example between text and non-text. You know EC is drawing but any EC is made of hierarchy of two-port networks, so it is like system of equations written in unusual for mathematics way. You can add or erase any element or change any nominal and recalculate a new circuit instantly. If you have a friend working in an electronics, you can give him any valid EC and ask him what he gonna do with it. His answer will be reading, not just looking at.

Re: Always Bet on Text (2014)

#335

Post by eimrine »

mantap wrote: post #203
soapdog wrote: post #196

It really depends on what you define as text. Does it involve special notations such as math? Can it involve diagrams? If all you mean is prose, as writing sentences in whatever native language you speak, then I must respectfully disagree with the OP. Yes, text is marvelous. I'm an author of multiple books and have a passion for the written word, but there are things for which there are better ways to convey informat…

I would draw the line like this. Text is anything you can represent using Unicode. Source code is text. Emoji are text. Some simple math expressions are text, but more complicated ones involving specialist layout rules are not. Architecture plans, no way.

Unicode answer is not so great because you can express compiled program in Unicode which obvious is non-text. Math expressions are always text because they always fit TeX or LaTeX.

Architecture is a beautiful example which provokes thinking. From first view it is a non-text, but an architect Antoni Gaudí has developed an attitude with hanging models which made his architecture plans to share some abilities of text even not being written. You may tell me I am crazy but his hanging models is a WYSIWYG model of something from real world so I consider it a text despite it could not be simply copied. IDK maybe a modern architect software shares that attitude, for make these models be copied easy. So my opinion is that architecture plan may be text at least from a structure point of view.

Re: Always Bet on Text (2014)

#336

Post by gowld »

BeetleB wrote: post #299
LordDragonfang wrote: post #286

Earlier quoted context omitted.

>Also very common to have multiple people conclude very different things from a photo. But that's true for text, too. The US just had a very major politicized event where two sides had two wildly different conclusions over what one man's words meant. And there's a grand legal tradition of taking one specific document and interpreting what its words mean and applying it to a plethora of situations. There's plenty of p…

The ideological message in 1984 is anything but clear when you compare it to what the message would look like had George Orwell merely written a nonfiction treatise, without resorting to any literary devices. If you know nothing of Orwell or the historical importance of 1984 and you read the book, it certainly wouldn't be clear that Orwell was trying to make a point, let alone what that point is. Some people would gu…

I disagree on obfuscation, and I think Orwell would too.

Orwell probably knew that if he wrote a dry academic tome, no one would read it, and that would obscure the message by default.

Re: Always Bet on Text (2014)

#337

Post by Banana699 »

atoav wrote: post #103
mihaaly wrote: post #63

Text is not the oldest - nor easiest to use -, text is a heavy abstraction born somewhat recently in human history. The oldest and easiest is the visual in 3D, then comes the visual in 2D. Those exist since we have eyes. Our brain have dedicated and sizeable infrastructure for that. Children can communicate in 3D (gestures, posture, expressions, objects) and in 2D drawings before learning textural communication with…

As someone who worked long in the visual field (Film, VFX, Graphics, Art) and has a ton of experience with text based expression (studied philosophy, programming) in my eyes picture based communication certainly is powerful. However reading pictures is much more subjective than most non-arts-educated might tend to realise. We had a weekly class where we would discuss scenes and pictures and what it evoked in people,…

The thing is, "text" and "pictures" are somewhat vague super-categories that aren't very well defined in a lot of comments on this thread. It's ironic that a discussion about the precision of text is noisy because of of the imprecision of text.

Are emojies text ? the blog post found a 4000-byte Twitter logo and cited this as evidence of the efficiency of text, but they forgot that there is a vast store of similary-sized photos that cost nothing more than 3 or 4 bytes and sometimes less, comparable to the very same letters they are using to write the post (and conveying much more than a single useless-on-its-own sound). By any definition that depend on the fact that text is stored as contiguous bytes, emojies ARE text.

Are 2D tables text ? they are not linear (not as we see them) yet one table with entries arranged appropriately could be much more percise and expressive than a truck-load of prose paragraphs. And an xy plot is much more expressive than both. 2D tables is not that much more expensive in storage than text though (a bunch of html tags surrounding the text is one representation), and an xy plot could be derived from a table by computation. So with a little bit of computation you gain the precision and expressiveness of a visual plot with the storage overhead of a bunch of numbers. This example (simplistically) mirrors current work in ML on compression of photos and videos. Imagine if computers are advanced enough that a picture is no more storage than an emojii, all arguments. about "efficiency" of text that appeal to its economic representation in computer memory become moot then.

The point I'm making is: "text" vs. "pictures" is not fine-grained enough of a distinction. The real distinction is [ "numbers" vs. "graphs" vs. "letters" vs. "tables" vs. ...), its a big multiway competition between various forms of symbols and symbol-organization schemas humans invented for a variety of purposes and arguments for or against any of them shouldn't appeal to how they are represented or communicated (text could be drawn on photos or arranged on powerpoint slides, images could be compressed and redrawn to a high degree of fidelity). And there is a lot of paradigms "in-between" that piggy-backs on one of the two representation but aren't either of them (plots, tables, emojies). Tools and technological limitations are also to blame for a lot of pictures' shortcomings, the ability to copy-paste being one of them.

A lot of discussions here would benefit alot if all parties state beforehand exactly what are their definition of text and what their argument appeals to in that definition.

Re: Always Bet on Text (2014)

#338

Post by dkersten »

Valgrim wrote: post #154
dkersten wrote: post #136

Earlier quoted context omitted.

> svg is text. If SVG counts as text, then text is just binary. The encoding is not important, the final result (an image you see on a screen, in the case of SVG) is. Besides, try reading out the actual SVG source text to someone and see if they can tell you what the image you are describing is. I dunno about you, but I find thinking in terms of paths and strokes to be rather meaningless. Or another example, do you c…

SVG is definitely human-readable. It's just drawing instructions, draw a line from there to there, then another one from there to there, etc. I can read it to someone over the phone with a pen and graph paper, and he can draw it exactly like it should be represented, with zero loss of quality. It's no different than someone looking at list of cash transactions. Numbers are text too. Of course, to extract meaning from…

> It's just drawing instructions, draw a line from there to there

That doesn't make it human-readable. I challenge you to tell me what (even the basic shape) this SVG is without loading it in a viewer (taken from a real SVG, not something handcrafted to prove you wrong, but the first thing I found online while searching for public domain SVG files):

https://gist.github.com/danielytics/90446a2314c1ad87017233eb...

Good luck figuring out what that is, even with graph paper. And that's a single layer, single (even if a bit complex) shape, single colour SVG. They only get more incomprehensible from here.

Re: Always Bet on Text (2014)

#339

Post by Lichtso »

smoe wrote: post #74
montebicyclelo wrote: post #13

> text is the most powerful, useful, effective communication technology ever, period. On the other hand, many (technical/mathematical) concepts are more effectively explained using diagrams/images. E.g. the very visual approach taken in Mathologer videos [1] makes difficult, esoteric, mathematical ideas accessible to a wider audience. [1] https://www.youtube.com/watch?v=iJ8pnCO0nTY

Also where does text end and diagrams/images start? E.g where does musical notation stand? For an untrained eye it is just a bunch of lines, dots and squiggles but others will read it like text and much more efficiently so than if it were written words. Similarly the example about the Wikipedia text of human rights being supposedly impossible to convey via image. But what does the text convey to you if you don’t spea…

> Also where does text end and diagrams/images start?

Text is the written form of language. And language is defined as a sequence of symbols. Thus, text is a subset of diagrams. Arbitrary diagrams can be of any topology (e.g. DAGs or even have cycles), however text always is of sequential topology (or a hierarchy as there is an isomorphic mapping to sequences).

Basically, if there is a canonical way of reading it out loud, then it has an explicit sequence and is thus text.

By that rule it follows that: Musical notation is text as it is strictly sequential (in time) and maybe hierarchical if there are multiple tracks. Most mathematical notations are text, as they are also sequential, even matrices are only a hierarchy of sequences which in turn can be serialized. Chemical / molecular diagrams are not text, they can contain cycles. Same goes for electronic and logic circuit schematics or rail way diagrams found in public transportation.

> Similarly the example about the Wikipedia text of human rights being supposedly impossible to convey via image

Schematics and diagrams are only use full if / because they are subject to a certain defined visual vocabulary. So if you had a visual vocabulary in that domain then you could express them as a diagram.

Re: Always Bet on Text (2014)

#340

Post by Banana699 »

VoodooJuJu wrote: post #157

The gist of what he's saying is pretty sound, but he should have gone deeper. For a deeper argument on the importance of "text", the written word, exposition - I recommend reading Postman's Amusing Ourselves to Death . People here are suggesting that images/video are superior, but there's the cliche, "a picture is worth a thousand words". Well, which thousand words? Do they convey the same thousand words to you as th…

Text is free of dominance gestures and emotions ? Have you ever paid attention to the words a politician might say? or the lyrics of a rap song? there are words in text that convey more emotions and gestures than if the speaker was face-to-face spitting the words personally on you. Even more when you know the original speaker beforehand and your brain is animating the syntax with their face and voice.

It's just a tradeoff ladder. Pictures represent pure unmitigated creativity, photographs represent the world, text is pictures + grammar, mathematics is when you compress the grammar to a few dead-simple rules that yiu could teach a bunch of silicate rocks to understand. At each step in the ladder, you lose freedom and expressiveness broadness but you gain freedom and expressiveness in a much more Domain-Specific way. And off course there are imperfections and leakages everywhere.

Post reply on HN