5 ms·

The author may well be on to something but personally I hate "story mode" in popular science books, and would really hate it in actual science papers, both the

by gumby 3y ago

The author may well be on to something but personally I hate "story mode" in popular science books, and would really hate it in actual science papers, both the ones I read for work and the ones I read for fun. I want to go straight to the equations -- often I prefer them to the graphs. But (not joking here) this is a perfect opportunity for an LLM -- two opportunities, actually. LLM A takes a dry paper and gives it context. It could make up the context but a good one would look up and offer an anecdote from Riemann's life or something. I see nothing wrong with that. And LLM B could take a paper with that stuff, which to me is fluff, and strip it all out, leaving the dry bones for me to pick over and savour. It would really just be another form of language translation, if a higher level one.

082349872349872 3y ago
I would say "story mode" in papers is like dancing about your doctorate: really cool when the combination works, but the latter is where all the value lies, so it shouldn't sacrifice anything for the former. What about LLM C, which takes a set of papers as vertices, forms an abstract simplex of all their combinations, and then spits out new papers on the ten most interesting higher-dimensional faces?
somenameforme 3y ago
This was my initial reaction as well, as I have complete disdain for the lowest common denominator approach to many things in modern society. But then something occurred to me - IMO one of the most well written scientific papers is Einstein's special relativity paper. [1] But it's absolutely a 'story mode' paper! A moderately educated individual could easily understand and follow the paper, even if they might not necessarily follow all the math. It just flows inordinately better than most modern papers - most of which are written on comparably simple and evolutionary (rather than revolutionary) topics. Of course this may be an issue of domain. I'm mostly interested in cosmology/astronomy/physics, where math is a tool rather than the object of the paper itself. [1] - https://www.fourmilab.ch/etexts/einstein/specrel/specrel.pdf https://www.fourmilab.ch/etexts/einstein/specrel/specrel.pdf
- Someone 3y ago
  I don’t see this article argument for a “lowest common denominator” approach. For example, how is replacing “Let M be a complete Riemannian manifold, G a compact Lie group and P → M a principal G-bundle.” by “One of the main problems in gauge theory is understanding the geometry of the space of solutions of the Yang–Mills equations on a Riemannian manifold.” doing that? It gets rid of “Lie group” and “G-bundle”, but adds “gauge theory” and “Yang-Mills equations” Also, “Of course, one should not give a detailed blow-by-blow account of every pitfall and wrong turn”, IMO, is an argument against doing that. It more or less says: “assume that there are steps your readers can make on their own”.
  - sethhochberg 3y ago
    As someone who never took much advanced math or physics in school and really doesn't understand what either representation of that material is staying, personally I find the second example far, far more approachable because it is far more googleable! If Gauge Theory is a concept required to relate to the other content in that sentence I've got no shot of knowing gauge theory is involved in the first example. I don't know what the arrow between P and M means. I'd have to look up what a G-bundle is. Its basically not clear to me which parts are syntax and which are proper nouns. The example which reads more like prose than an equation expressed in terms of English words gives me much, much more context for where to begin reading about what I don't know.
  - gumby 3y ago
    One important element for me is that the framing should be as specific as possible. The article's example, regrettably isn't the best example of this, but: by saying "one of the main problems..." the author (quite reasonably, not click-baiting) is trying to say "this paper is about something important and is worth reading, or at least reading the abstract". Unfortunately for me (not necessarily others) these kinds of context act as framing, so I am less likely to match to a similar case in a different domain. Here's a CS example: let's just say you might have found a way to, say, compress the TLB or use fewer instructions to use it which speeds things up in most cases but slows them down in a few corner cases. You could start by talking about the problem of paging systems under high load or large RAM or something -- great! But if you described a novel hashing architecture, later in the paper pointing out that it's "..useful, for example in a pager", I might read the paper and say "holy cow this would work well for this thing I'm working on". That's why I prefer just the dry bones. I can hang whatever flesh I want onto them. But I know not everybody is like that, and that's OK. The world isn't supposed to pander just to my need (though it should, dammit!)
gravescale 3y ago
I think a lot of it comes from following advice similar to "Write a Catchy First Paragraph" and it goes too far. You end up starting out with bizarre barely-sequiturs like "Fiona was a graduate student in lower New York in a family cafe run by a man named João sipping her usual order of single-origin cappuccino on a rainy Wednesday" before we even find out what the article is about, let alone what the actual insight is. Furthermore, a lot of popular science ends up using the people involved as the lens through which the ideas are eventually viewed. Which makes a lot of sense for professional writers who are probably more attuned to the human interest than technical people. For an example, the first thing I did with a Lego vehicle model kit was to throw the little figurine into the "junk bits box" and proceed with a now-robotic model. Many things are more like Oppenheimer than Trinity Device Annotated Systems Manual. Which doesn't mean they're wrong, per se: the audience for it is probably bigger and the overall "utility" of the work is higher. And even a grump like me knows you shouldn't completely ignore human factors. On top of that, people who can write the complex technical stuff often don't want to mess about in the middle ground. But the bimodality is still annoying to me: people-centric "Stories" or deeply-involved dessicated technical material that I don't easily understand if it's not my field and not so much in-between.
darby_eight 3y ago
> leaving the dry bones for me to pick over and savour. How would you understand the relevance the equations have to the overarching finding of the paper? Narrative is just as important in tying together apriori reasoning as it is in other contexts in all but the most trivial findings, and much of computer science is not, in fact, apriori, requiring argumentation to justify the abductive reasoning within.
nextaccountic 3y ago
> The author may well be on to something but personally I hate "story mode" in popular science books, and would really hate it in actual science papers, both the ones I read for work and the ones I read for fun. I want to go straight to the equations -- often I prefer them to the graphs. I think John Baez agrees with you. At the conclusion he says, > The ideas here take practice to implement well, and they should not be overdone. I’m certainly not saying that a good math paper should remind readers of a story. Ideally the tricks I’m suggesting here will be almost invisible, affecting readers in a subliminal way: they will merely feel that that paper is interesting, carrying them in a natural flow from the title to the conclusion.

by gumby 3y ago

082349872349872 3y ago

I would say "story mode" in papers is like dancing about your doctorate: really cool when the combination works, but the latter is where all the value lies, so it shouldn't sacrifice anything for the former. What about LLM C, which takes a set of papers as vertices, forms an abstract simplex of all their combinations, and then spits out new papers on the ten most interesting higher-dimensional faces?

somenameforme 3y ago

This was my initial reaction as well, as I have complete disdain for the lowest common denominator approach to many things in modern society. But then something occurred to me - IMO one of the most well written scientific papers is Einstein's special relativity paper. [1] But it's absolutely a 'story mode' paper! A moderately educated individual could easily understand and follow the paper, even if they might not necessarily follow all the math. It just flows inordinately better than most modern papers - most of which are written on comparably simple and evolutionary (rather than revolutionary) topics. Of course this may be an issue of domain. I'm mostly interested in cosmology/astronomy/physics, where math is a tool rather than the object of the paper itself. [1] - https://www.fourmilab.ch/etexts/einstein/specrel/specrel.pdf https://www.fourmilab.ch/etexts/einstein/specrel/specrel.pdf

Someone 3y ago
I don’t see this article argument for a “lowest common denominator” approach. For example, how is replacing “Let M be a complete Riemannian manifold, G a compact Lie group and P → M a principal G-bundle.” by “One of the main problems in gauge theory is understanding the geometry of the space of solutions of the Yang–Mills equations on a Riemannian manifold.” doing that? It gets rid of “Lie group” and “G-bundle”, but adds “gauge theory” and “Yang-Mills equations” Also, “Of course, one should not give a detailed blow-by-blow account of every pitfall and wrong turn”, IMO, is an argument against doing that. It more or less says: “assume that there are steps your readers can make on their own”.
- sethhochberg 3y ago
  As someone who never took much advanced math or physics in school and really doesn't understand what either representation of that material is staying, personally I find the second example far, far more approachable because it is far more googleable! If Gauge Theory is a concept required to relate to the other content in that sentence I've got no shot of knowing gauge theory is involved in the first example. I don't know what the arrow between P and M means. I'd have to look up what a G-bundle is. Its basically not clear to me which parts are syntax and which are proper nouns. The example which reads more like prose than an equation expressed in terms of English words gives me much, much more context for where to begin reading about what I don't know.
- gumby 3y ago
  One important element for me is that the framing should be as specific as possible. The article's example, regrettably isn't the best example of this, but: by saying "one of the main problems..." the author (quite reasonably, not click-baiting) is trying to say "this paper is about something important and is worth reading, or at least reading the abstract". Unfortunately for me (not necessarily others) these kinds of context act as framing, so I am less likely to match to a similar case in a different domain. Here's a CS example: let's just say you might have found a way to, say, compress the TLB or use fewer instructions to use it which speeds things up in most cases but slows them down in a few corner cases. You could start by talking about the problem of paging systems under high load or large RAM or something -- great! But if you described a novel hashing architecture, later in the paper pointing out that it's "..useful, for example in a pager", I might read the paper and say "holy cow this would work well for this thing I'm working on". That's why I prefer just the dry bones. I can hang whatever flesh I want onto them. But I know not everybody is like that, and that's OK. The world isn't supposed to pander just to my need (though it should, dammit!)

gravescale 3y ago

I think a lot of it comes from following advice similar to "Write a Catchy First Paragraph" and it goes too far. You end up starting out with bizarre barely-sequiturs like "Fiona was a graduate student in lower New York in a family cafe run by a man named João sipping her usual order of single-origin cappuccino on a rainy Wednesday" before we even find out what the article is about, let alone what the actual insight is. Furthermore, a lot of popular science ends up using the people involved as the lens through which the ideas are eventually viewed. Which makes a lot of sense for professional writers who are probably more attuned to the human interest than technical people. For an example, the first thing I did with a Lego vehicle model kit was to throw the little figurine into the "junk bits box" and proceed with a now-robotic model. Many things are more like Oppenheimer than Trinity Device Annotated Systems Manual. Which doesn't mean they're wrong, per se: the audience for it is probably bigger and the overall "utility" of the work is higher. And even a grump like me knows you shouldn't completely ignore human factors. On top of that, people who can write the complex technical stuff often don't want to mess about in the middle ground. But the bimodality is still annoying to me: people-centric "Stories" or deeply-involved dessicated technical material that I don't easily understand if it's not my field and not so much in-between.

darby_eight 3y ago

> leaving the dry bones for me to pick over and savour. How would you understand the relevance the equations have to the overarching finding of the paper? Narrative is just as important in tying together apriori reasoning as it is in other contexts in all but the most trivial findings, and much of computer science is not, in fact, apriori, requiring argumentation to justify the abductive reasoning within.

nextaccountic 3y ago

> The author may well be on to something but personally I hate "story mode" in popular science books, and would really hate it in actual science papers, both the ones I read for work and the ones I read for fun. I want to go straight to the equations -- often I prefer them to the graphs. I think John Baez agrees with you. At the conclusion he says, > The ideas here take practice to implement well, and they should not be overdone. I’m certainly not saying that a good math paper should remind readers of a story. Ideally the tricks I’m suggesting here will be almost invisible, affecting readers in a subliminal way: they will merely feel that that paper is interesting, carrying them in a natural flow from the title to the conclusion.