Summarizers start with the default assumption that reading is an obstacle standing between the reader and some kind of reward. Even the idea that knowledge is something that is capable of being transferred is something that has to be assumed at one’s peril.
On the other it's those of us who’ve read in the old school style, for fun, in private that are more convinced of the opposite than anyone. If anything getting summaries might be the worse of both worlds because one might be left with the false impression of understanding where there is none.
Anyways, as was pointed out elsewhere in the thread, even English majors and other serious literary people often have no idea what they’re talking about, which just goes to show that people who were going to read will do it regardless of what else is happening in their life, and people who weren’t going to read will not read even if it’s their major. In this sense, LLMs don’t really change anything. The same person operating the tool will continue to be the same person in either case.
I read for work, but I also read in my spare time. I love reading about things that I know very little about. Books still generally live up to their synopsis and respect your time, if you choose them well. I mostly stick to books for my leisure reading.
Long-form articles have become like opening a box of chocolates in the Forest Gump sense. "You never know what you're going to get." That half-nonsensical title that somehow got you to click isn't going to be explained, clarified, or elaborated on until you're fifteen minutes in, and its a coin-toss if the article will even answer the questions it pretended to ask. The odds are high that the author will go off on a tangent and never return.
When you're baited into reading a rambling, unfocused longform article that has nothing to do with it's title, it often feels like you've been swindled out of your time. That's because you have been swindled. I heartily encourage people to use AI to produce abstracts of long-form articles before reading them. It's like installing an alarm system. Don't let long-form thieves steal your time.
And it's the New Yorker that is frequently a culprit here. Too many articles talking about the journey the reporter went on to write their article. The low signal to noise ratio is a decent chunk of the reason I unsubscribed. Too few articles paid off.
But this isn't new, this is something that was very common pre-internet too.
It's an art form, and it's about exploring the people behind the ideas as well as the ideas themselves.
I used to love reading the Saturday and Sunday magazines that came with the paper, this was back in the 1990s. Many of those were always of this long form, rambling, structure.
If you can see the pattern it's obvious once you start reading.
There's a sort of teaser at the beginning. Then they dive into a person's history like this:
Super Bionics is an exciting new form of prosthetic that's revolutionising lives.
Steve Jones was sitting on his porch when he first thought of super-bionics. Steve had always been fascinated by robotics when he was growing up. At 3 years old.........
5 paragraphs later they'll finally do a bit more about the super-bionics.
Then each section gradually moves the story forward in exactly the same way. They intro the subject. They introduce a new person behind the subject and explore them and their motivations. Then they say a bit more about the actual subject.
You're mistaking a writing style with time wasting.
I don't have the patience for it any more, but lots of people do.
I just watched Wild Ones on AppleTV, and I feel the same way about that series. IMDb has a tag line "Investigates the delicate ecosystems of our globe and finds information on how to help conserve and protect the most priceless endangered species." However, the entire thing felt more like a glorified influencer vlogging their vacay. It was much more about the camera people than the animals. I know this wasn't Planet Earth, but the footage they acquired was not the prominent bits of the series. It was produced well enough that I watched each episode and I was curious about each episode, but they all left me with the same feeling of meh about it.
Some of these articles are definitely more about the author being able to say they write than it is shedding light on anything or providing any kind of insight. It's all just gross to me
In this sense, LLMs don’t really change anything. The same person operating the tool will continue to be the same person in either case.
I don't understand why more people don't get this. I've told everyone who will listen in my org that implementing LLM's isn't going to solve the problem of people wasting our time reaching out to us with questions that have already been answered in our KA system. If someone was going to type something into an LLM, they could have typed it into a search bar. People don't skip the documentation because they can't find it; they skip because they don't want to read! They want to bug a live a person (and make it their problem)!
I was correct. Now we have a costly LLM implementation and have our time wasted with questions that are already answered.
You put the LLM in the wrong place. Put it as the first tier of email reply, then the people reaching out to you will get their LLM reply and not bug you.
You must be the org that operates this mysterious, legendary KB system that has any actual information in it.
None of the KB systems I've ever encountered, from orgs big and small, reputable or otherwise, had in them any useful information whatsoever. It's all just bullshit, roundabout sales, and a lot of answers to questions no actual human being would ever ask.
Whether or not to put LLM in the KB's search bar is immaterial. Myself, when I see a KA syste, I immediately close the tab and "bounce" back to the search engine.
> People don't skip the documentation because they can't find it; they skip because they don't want to read!
Yes, but also because they don't trust you to provide enough information to diagnose and solve their problem. There's no point wasting the time when experience tells you most customer-facing KBs are nothing but false hope and misleading headlines.
> They want to bug a live a person (and make it their problem)!*
Yes, but not because they hate that random support person - it's because it's the only remotely reliable way to solve the problem itself. Companies shouldn't complain, not after standardizing knowledgebases and phone menus, which are all implicitly and often explicitly designed to keep customers away from support staff - and instead of providing a solution, they're optimized to make the user think the solution exists somewhere and they're just too dumb to find it (therefore, user's own fault, not company's fault; customer's brand perception unaffected).
As for making it the support person's problem - that's literally their job. That's what they're paid for.
Another part of what happened is that the comment section feels more succinct and insightful than the actual article. Articles have to be long form, comments get to the point. It's sort of like your comment is the LLM response I wanted all along. And now we can personalize our reading and have a more meaningful outcome.
Maybe long form content solved a need back in the day when things were printed on paper and figured out well in advance, crossing their fingers on the relevance, and with where we are now we can suss it out without all the reading-as-middleman-to-knowledge
It seems like you're saying almost the exact opposite of the person you're responding to.
"Reading" an article through its comments makes the assumption that those commenting actually read and understood the article. This seems like a risk though, as there is an entire ecosystem of people who are just knowledgeable enough to be listened to by those with the same or slightly less knowledge of the content or field.
How many times have you sent a meme or made a referential comment about some piece of media that you've never even seen? Big Lebowski, Breaking Bad, and American Psycho memes are completely intelligible across the internet even though many people have never actually seen them.
I think the argument of the person that you're responding to is that these dilettantes would exist regardless of the tools that were out there, LLMs or otherwise. There have always been people that prefer to talk about things than to read and consume them.
The assumption that long form content is a relic and that reading is no longer necessary for knowledge seems absolutely crazy to me, but it does seem to be a common enough mindset that I've run into it with students that I mentor. It seems logical to me that if you could learn something in one hour, then by definition your knowledge in that subject can not be deep. But it seems like there are plenty of people that I work with and talk to that think a crash course or podcast is all you need to be an expert in something.
I think this comment mistakes “understanding the plot” as the main goal of reading, but misses that reading (as a process, a verb) can be the goal in itself, at least in terms of recreational reading. Summarization misses all that experience, just like reading the synopsis of a movie isn’t the same as viewing the art. I don’t want everything in my life to be just a rush to the ends, anymore than I’d want to trade the human experience of hugging my child to be reduced to simply understanding “an increase in reading oxytocin creates bonding leading to higher resource investment and survivability.”
A rush to “get to the point” when dealing with art feels very much like the tech-obsessed productivity porn that can miss the forest for the trees.
> "It's sort of like your comment is the LLM response I wanted all along."
A passage from E. M. Forster's "The Machine Stops" springs immediately to mind.
"... 'Beware of first-hand ideas!' exclaimed one
of the most advanced of them. 'First-hand ideas do not really exist. They are but the physical impressions produced by love and fear, and on this gross foundation who could erect a philosophy? Let your ideas be second-hand, and if possible tenth-hand, for then they will be far removed from that disturbing element — direct observation. Do not learn anything about this subject of mine — the French Revolution. Learn instead what I think that Enicharmon thought Urizen thought Gutch thought Ho-Yung thought Chi-Bo-Sing thought Lafcadio Hearn thought Carlyle thought Mirabeau said about the French Revolution. Through the medium of these ten great minds, the blood that was shed at Paris and the windows that were broken at Versailles will be clarified to an idea which you may employ most profitably in your daily lives. But be sure that the intermediates are many and varied, for in history one authority exists to counteract another. Urizen must counteract the scepticism of Ho-Yung and Enicharmon, I must myself counteract the impetuosity of Gutch. You who listen to me are in a better position to judge about the French Revolution than I am. Your descendants will be even in a better position than you, for they will learn what you think I think, and yet another intermediate will be added to the chain. And in time' — his voice rose — 'there will come a generation that had got beyond facts, beyond impressions, a generation absolutely colourless, a generation seraphically free From taint of personality, which will see the French Revolution not as it happened, nor as they would like it to have happened, but as it would have happened, had it taken place in the days of the Machine.'
Tremendous applause greeted this lecture, which did but voice a feeling already latent in the minds of men ...
"
Or perhaps what Terry Pratchett wrote about the river Ankh may apply: "Any water that had passed through so many kidneys, they reasoned, had to be very pure indeed." One has to wonder if people are thinking "Any idea that has passed through so many layers of minds has to be thoroughly refined indeed."
I mean, the insurmountable problem is - and will always be - that true knowledge lies beyond words. You can communicate and articulate and pontificate and these are all good things, but even at their best efforts they will never be more than a mechanical process that will never quite get you there. In other words, there will never be the right words to “get it” because what there is to get is fundamentally unexplainable.
It’s like trying to explain what one may see hear or feel when their on vacation at an exotic new location by talking about the train tracks that brought you there.
So when you’re reading you’re not downloading packets that add up to some kind of point. Instead, in the absolute best case scenario, you’re simulating the experience, according to the author’s recommended doses, of someone else “acquiring” knowledge. This “someone else” is the nameless reader the book was written for but they are not you.
> people who were going to read will do it regardless of what else is happening in their life, and people who weren’t going to read will not read even if it’s their major. In this sense, LLMs don’t really change anything. The same person operating the tool will continue to be the same person in either case.
This seems extremely detached from reality. Just to give you an anecdotal example, I used to love reading books as a kid, but you'd be extremely hard pressed to find me reading one now. Clearly my reading habits have changed, and so it cannot be some intrinsic property written in my fate (which seems to be assumed to exist in the quoted framing).
Conversely, book purchases wouldn't fall or rise, especially over long timeframes, if there weren't changes in reading habits. It just doesn't make sense to portray people as these immovable objects, whose desire to read has been inscribed into them at birth.
> Summarizers start with the default assumption that reading is an obstacle standing between the reader and some kind of reward.
When reading novels for pleasure is not among your hobbies, it most likely will be. You bet I don't often want to read e.g. documentation: I just want to get my hands on the magic incantations or magic phrases / values required, and move on with my day.
But this is true even for literature class, with its mandatory readings. I was one of the more naive folks in my class, so I'd make an earnest attempt at reading through those things. At the end, that only mattered to the extent that I can now tell you it was a complete chore. And the "reward" at the end of those was good test scores of course - something others could replicate by just relying on, then manually written, summaries.
> one might be left with the false impression of understanding where there is none.
It is at most only an opinion that there is no understanding there, and so is that the impression the person would have about their understanding is false. Understanding is not a binary property, despite what e.g. the anti-AI crowd would make one believe, and you're no mind reader. I think that's pretty agreeable at least.
I found very quickly when I started associating with philosophy majors, reading the material usually does not correlate to understanding it, and many people simply don't want to put in the effort to understand something that isn't clear at first glance.
I do think your point that some people wouldn't read regardless is bad. People will waste time doing things that are available to them. If less forms of entertainment are available, picking up a book becomes more commonplace.
I often wonder how much humanity's rate of progress will slow (or regress) in the coming decades.
A lot of modern business books are guilty of word stuffing, like webpages trying to hit 500 words to get SEO ranked. Dense works and older books tend to have a better reading journey.
- it's possible to transfer knowledge, as demonstrated by the fact that human civilization exists. It's not always easy, doesn't always succeed, and reading is a part of that, but it's possible and happening. I'm confused about your intended meaning in claiming otherwise.
- It's very difficult to distinguish between (especially one's own) understanding and a false impression thereof. To an overwhelming degree, the main realistic way is applying the knowledge, which is easiest when far removed from the activity of reading.
- One's upbringing, environment, social circle, etc., strongly influence one's propensity for reading, both for work and for pleasure. People change, especially as long as they're young, but even adults do in a major way according to conditions.
Well, education researchers are trying to communicate that knowledge is never "transferred".
IIRC (I don't recall where I read about this), there are two problems:
1. "transfer" gives the impression that a person can copy their knowledge to another person, but that is not the case. The teacher says, writes the words or even demonstrates, but the brain in the student is making its own connections and tries to explain it in its own frame of reference. It may click, or may not, or may even click in the wrong way, leading to learning a different lesson from the one being taught.
2. The teacher may have tacit knowledge they do not know they have to teach, or convey by some other means. Most teachers don't even realize that this tacit knowledge is not present in their students.
So, maybe nitpicking a bit, but "transfer" is not the right word for it.
The pace of advancement in human civilization, expecially in science and technology (and all the conveniences and economic multipliers that have resulted) was very slow until literacy became widespread.
Before most people could read, you would learn a trade from your father or as an apprentice. Knowledge was handed down but you pretty much learned "the way it has always been done" and improvements were slow.
Once we all could learn from books and publish our discoveries, the spread of knowledge and the pace of advancement exploded. We went from farming with animal labor to walking on the moon in under a century.
I had this exact thought the other day. Social media is information porn. Endless amounts of empty information that gives you the feeling of acquiring knowledge, with none of the substance.
Doesn't it really depend on what we're reading? I generally wouldn't want to skip any words in a fiction novel. I would love to skip words in most self-help books that turn out to be 3 paragraphs of "the point" and 100-200 pages of fluff to get the book on the shelf. Another perfect example is nearly any online recipe with 25 screens of fluff at the top to get to the 1 screen recipe.
A great perspective but one problem remains: AI will radically change the book market. Great new books will be even harder to find as we are drowning in a sea of words. How do we stay afloat?
Seek out book enthusiasts, which in some ways (tools, internet), is easier than ever, and find your tribe(s) that align with your taste. AI or not, the volume of books is ever increasing. Word of mouth, personal recommendations, curated lists, all are only increasing in value.
> If anything getting summaries might be the worse of both worlds because one might be left with the false impression of understanding where there is none.
It's bad for the individual, but even worse for the collective. The AI summary reader isn't just convinced they understand, they also share that incorrect understanding in discussion. They effectively inject LLM slop into real life conversations and forget the subjectivity of reading.
This leads to what I consider more harmful. Discussions where the particiants themselves don't even believe in the stuff they are arguing. Where human beings, devoid of their own subjectivity, sling summaries and empty "facts" at each other. As if what 1984 textually said is important in any way beyond how you and I, the humans, connected with it.
i for one always enjoyed reader's digest even though i knew it was heavily critiqued by all the fancy people.
i still like reading the adapted books into chapters from the walrus too. i havent been a subscriber in a while, but they tend to be nice reads that dont require me to commit to reading the same thing for the next couple weeks on the toilet
I'm not sure what's happening, but I am sure it isn't new.
I had to learn the hard way 15 years ago that the average American adult cannot parse a full-page email in standard English. It seemed crazy to me at the time and seems crazy to me now, given that the average adult has completed elementary school, but most people are barely functionally literate at all.
I don't expect you to believe me. It's a weird claim. But walk into any average grocery store and hand someone a page out of a book and ask them to read it out loud to you. Many people are so aware they can't that they will refuse to try. Of the ones that do, you will struggle to find one who can read the text with anything like the fluidity or inflection they would use to speak the same words. If you give them time to prepare, they'll probably be able to get through it in a few minutes, but nobody's putting that kind of effort into a text-only email, even if it's important for work.
Reading is so difficult as to be a chore for the average person. They don't just see written words and know what they say. They really have to work to get meaning out of written text.
With the proliferation of other means of taking in information, many of which require no effort of any kind beyond hitting play and staying within earshot, why would people choose to read? They didn't want to do it before. And now they don't need to do it either.
> walk into any average grocery store and hand someone a page out of a book and ask them to read it out loud to you. Many people are so aware they can't that they will refuse to try. Of the ones that do, you will struggle to find one who can read the text with anything like the fluidity or inflection they would use to speak the same words
Of course.
I would likely find this situation very unsettling, if not stressful. I would probably be caught off guard. I would probably perform badly if I accept at all for all sorts of reasons unrelated to my ability to read smoothly. Unless I'm feeling particularly positive and gather the energy to pace my mind, apply myself, enter a character like I'm having a role in a play, and remember to be slow, and to forget about the content. Reading out loud is more complicated than simply reading for yourself: you are, at the same time, both reading and articulating speech.
What's more, when you speak, you are using your own words, your own (oral) phrasing, and you know where you are going, or at least enough to have a decent prosody. When you read out loud, you are reading phrasing from someone else, which may make the prosody less smooth. And if you haven't read the whole sentence in advance, or at least ahead enough, you may struggle.
Independently of knowing how to read (fast), reading out loud probably takes practice to be smooth.
With your experiment, you are testing all sort of things unrelated to knowing how to read, to the point that you can't draw any conclusion on the ability of people to read.
Parsing a text also depends on the writer's ability to write well. If the text is boring or its phrasing is overly complicated, yeah, it will be difficult to read.
We also live in a world plagued by focus disruptions. You are not only dealing with people's ability to read, but also their ability to remain focused… in a setting where they possibly get interrupted very frequently.
For all sorts of reasons, reading a wall of text is indeed hard. This includes the reader's environment, the presence of their smartphone next to them, how tired they are, whether they are concerned by something else, whether the text is actually interesting to them, and how well the author of the text writes.
> With your experiment, you are testing all sort of things unrelated to knowing how to read, to the point that you can't draw any conclusion on the ability of people to read.
Even "caught off guard", not "feeling particularly positive", and "reading phrasing from someone else", some people have so fundamentally mastered reading as a skill that this wouldn't be difficult for them even if the challenge were attempted drunk. The point of the parent comment was, I believe, that for such people it may be hard to imagine that others would believe "reading a wall of text is indeed hard" because it isn't for them.
Do schools not ask kids to read out loud in class anymore? Any one of my classmates could read a page out loud fluently even if you stopped them randomly at the grocery store, and most of us are not even 30 yet.
It's actually not that hard to speak text fluently. You need to practice, but only a little. The trick is to speak a little slower than you read, and pre-emptively flick your eyes to the right, skimming the oncoming text, so you have a buffer of context in your mind. Pause a little more at sentence boundaries; this gives you more headroom and makes you sound more refined. The idea isn't to fully process the future text, but to get a rough sense of where it's going -- a question, a subordinate clause, a list, etc -- so you can adjust your intonation in advance.
As an experiment I spoke your comment aloud as I read it. It sounded natural and I made no mistakes at all.
You have spent a lot of time and effort to explain precisely how and why you failed to get GP’s point, which is completely true and valid despite whatever unintentional misdirection you have posted in response.
I think you're missing the point that most of the things that make this hard for people are also what makes it hard enough for them to read that they won't do it well, even for things like emails.
Reading is a skill, and it's not the same as being (technically) literate. For someone who's a competent reader, there's nothing at all more intimidating about being asked to read a page from a book (you can choose the book, but a page from a book is not a "wall of text," and the book's publication also implies at least some pre-screening for accessibility) than being asked to have an impromptu conversation about where to find the potatoes, and whether you know any tasty recipes for them.
Most people don't read well. They read well enough to understand street signs before they pass them, or to get through an article intentionally written to a 6th grade level. But anything where the details are important? Or that requires explanation? They either cannot do it, or find it so difficult that they won't.
>I don't expect you to believe me. It's a weird claim. But walk into any average grocery store and hand someone a page out of a book and ask them to read it out loud to you. Many people are so aware they can't that they will refuse to try. Of the ones that do, you will struggle to find one who can read the text with anything like the fluidity or inflection they would use to speak the same words. If you give them time to prepare, they'll probably be able to get through it in a few minutes, but nobody's putting that kind of effort into a text-only email, even if it's important for work.
How is that relevant ? Reading out loud is not the same as reading ? I struggle to keep pace with reading out loud - because I am just not used to it. I do better now that I read to my kids, but I don't think I have reading problems - reading out loud is a skill that's on top of reading. Speech patterns, words you're used to pronouncing, etc. all make reading out loud more difficult than speaking or just reading IMO
I can read in my head fine. Reading aloud I'm slow and words come out stilted. It's a skill that takes practice to be good at, and it's rare to need. I don't think that's a useful metric.
The underlying truth here is worse than 'majority of educated are illiterate'.
Collectively, we've built these delusions into our culture. Perhaps there is less suffering this way.
I don't want to correlate the ability to read aloud to a stranger about the same as parsing. It's possible to read and not parse what you're saying (that's what teleprompters are for) and its possible to parse and not speak. Do we have formal studies for comprehension?
> Many people are so aware they can't that they will refuse to try. Of the ones that do, you will struggle to find one who can read the text with anything like the fluidity or inflection they would use to speak the same words
Is this tru’ish? I’m not refuting it I’m just a little shocked this might be the situation we’re in. I know generally people now struggle to consume long form content but it even being able to read a story?
It’s probably true insofar as if random assholes accost me and ask to read a page from a book I’m more likely to employ old Anglo Saxon than engage with their stupidity.
I read like 80 books in a slow year but if someone approached me at the grocery store and asked me to read even a children's book I would refuse. It's not because I can't read, it's because when I'm at the grocery store I am kind of busy getting my groceries and don't want to engage with a stranger who has an odd request.
That stranger might conclude I can't read, when in reality I devour books, am the beta reader for my two published-author friends, and probably edit the "thoughts on books I've read" section of my personal wiki more than any other.
A weirdo approaches you with a highly unusual request. You suspect the goal is to humiliate you or make you into a dummy somehow, which is actually their goal. No thanks.
> I know generally people now struggle to consume long form content but it even being able to read a story?
There is reading and then there is reading an unknown text out loud in public while being judged. The two are not the same thing. I had to read publicly when I was a student years ago when smartphones were a new thing. They handed me the text with instructions to reread it at least twice ideally out loud before performing.
The point is, it is not like 20 years ago a good student would be expected to read super fluidly out loud in public without at least a little preparation.
> I had to learn the hard way 15 years ago that the average American adult cannot parse a full-page email in standard English. It seemed crazy to me at the time and seems crazy to me now, given that the average adult has completed elementary school, but most people are barely functionally literate at all.
They can, they just won't, because they don't give a fucking shit. The moment you hit adulthood in modern times you're bombarded with bazillion of bullshit. Do you seriously believe your meticulously hand-crafted email is high enough on someones radar that they'll actually pay attention?
This. If someone sends me a full page email, I immediately skip it. No time for that. They'll follow up more concisely when it becomes important. Or if it requires such elaborate description, a phone call or meeting would have probably been a better channel for this communication. Likewise, anytime I start typing an email and it gets lengthy I know I'd be better off picking up the phone/scheduling a call.
Is it fair to assume that comfort in speaking/oration correlates to reading comprehension?
I don't know that what you've described is any different now than 20 years ago when I was in high school. People struggled to read aloud texts like plays or classic literature. I would use that as a bar for complex prose that benefits from good narration; often the point was to encounter unfamiliar words or meter and to read with the purpose of critical analysis. A friend of mine is an author and we did a group reading of a play they'd written. Quite hard to do off the bat. Similarly if you've ever tried to DM a role-playing game like DnD, where the text you read is semi-randomly chosen based on what the characters decided to do (also, will you role-play dialog?).
I've worked in academia for 10+ years at this point. You can tell almost immediately if a presenter has had practice, knows their material or is comfortable speaking in public. Lecturers and professors are, unsurprisingly, often very comfortable giving presentations and there are people who live for conferences and working groups. We're required to read dense material frequently. Understanding of a piece of work, or the attention span required to ingest a scientific paper, does not necessarily mean you could read it aloud fluently.
It's obvious if you think about learning a foreign language. Parsing written text silently is a completely different task than producing speech with the right intonations, even if you recognize the words.
Reading aloud fluidly requires a certain speed and anticipation. You have to be reading the next word in your head while you are still pronouncing the previous one. This isn't necessarily correlated with reading comprehension which is purely input.
I don't know. Talking with strangers is always kind of insane, or at least a pretty illogical leap of faith, because there's not much chance that the interaction is interesting or amusing. Now a stranger that walks up to me and demands some dramatic reading of Paradise Lost, stat? Ok, I'm intrigued, hold my beer and buckle up while I let you hear about regions of sorrow and darkness visible because the produce section has never seen some shit like this
I think one can possess decent reading comprehension skills while also be deficient in their reading aloud skill. Beyond classroom requirements, reading aloud is not a typical activity many people engage in even if reading silently is.
... if I was asked by a stranger to read a page of book for them out of blue in a store, I'd be staring dumbfounded, questioning whether everything's alright with them.
Interestingly enough Claude has me reading much more. Especially with math books, one of the greatest challenges to self-study can be making sure you are in fact getting the concepts correct. Without this it's easy to get fairly deep into a book only to give up once you realize you haven't quite built the picture in your head right. Often you do get it, but it takes multiple re-reads/alternate views of the problem.
With Claude as I read I can constantly check my understanding. When my response elicits a "Well, not exactly..." I know I have to go back. This combined with the ability to have Claude clarify formula details from a phone picture has rapidly accelerated my learning and has me reading much more these days.
Claude is also pretty good at subject specific recommendations, especially when you're looking for a specific type of treatment of a subject.
I saw an example recently of a sort of “AI Codec” :
A person has to send a message to a respected figure of authority. They organize their thoughts and requests into a clear and concise bulleted list with explanations. But, that seemed heavily informal and unprofessional. So they used AI to convert the bullets to paragraphs and sent it out.
The authority received the large body of text but, due to time commitments and attention, they didn’t have time to read it all. They used AI to convert the text to a concise bulleted list.
This is great. Maybe this hints at a different approach, instead of asking the LLM to expand the bullets into paragraphs we could ask it to generate a text that, when asked to be summarized by an LLM will produce the original bullets.
What's happening to reading? Followed by several popups that unaccountably take ages to make themselves known and prevent me from reading the damn article.
Not the point they wanted to make, but a point nonetheless.
Too many people writing is a problem they have not discussed in this article. Information is too diluted. I dont have the time to read 5000 books on forex trading. What I want is an LLM crunching all 5000 forex trading books into one book with very concise information. I could apply this idea to "Value investing", "Cookbooks", "Startups", "Business", "Finance" and so on. Does anyone here on HN have an LLM capable of doing this or working on something along these lines?
Six months ago there seemed to be a flood of people who wanted to normalize Dyslexia and were pitching startups that the 75% of people who can read just didn't need because... they can read.
Haven't seen so many pitches for summarizers lately.
there was software called Copernic Summarizer ~25 years ago that was so useful for taking huge articles and condensing them into a paragraph. I have no idea how it worked. At some point i lost access to several pieces of software i bought in that era, also including ambrosia software's catalog which i had purchased. I think i lost my gte email address or something, can hardly remember.
I haven't used chatgpt (or whatever) for summaries in a couple of years, so i have no idea what SOTA is; although "chat with a document" seems like it'd be more useful in general than a summary for the way i eschew long-form articles.
For me chat with a paragraph in a language I sorta know (Japanese) or wish I knew (Chinese) is really useful. I ask for a translation and see discordances with what I can read and ask about them and get good answers. I also can lean on translations from my text and insist that certain words get used, etc.
Readit News
On the other it's those of us who’ve read in the old school style, for fun, in private that are more convinced of the opposite than anyone. If anything getting summaries might be the worse of both worlds because one might be left with the false impression of understanding where there is none.
Anyways, as was pointed out elsewhere in the thread, even English majors and other serious literary people often have no idea what they’re talking about, which just goes to show that people who were going to read will do it regardless of what else is happening in their life, and people who weren’t going to read will not read even if it’s their major. In this sense, LLMs don’t really change anything. The same person operating the tool will continue to be the same person in either case.
I read for work, but I also read in my spare time. I love reading about things that I know very little about. Books still generally live up to their synopsis and respect your time, if you choose them well. I mostly stick to books for my leisure reading.
Long-form articles have become like opening a box of chocolates in the Forest Gump sense. "You never know what you're going to get." That half-nonsensical title that somehow got you to click isn't going to be explained, clarified, or elaborated on until you're fifteen minutes in, and its a coin-toss if the article will even answer the questions it pretended to ask. The odds are high that the author will go off on a tangent and never return.
When you're baited into reading a rambling, unfocused longform article that has nothing to do with it's title, it often feels like you've been swindled out of your time. That's because you have been swindled. I heartily encourage people to use AI to produce abstracts of long-form articles before reading them. It's like installing an alarm system. Don't let long-form thieves steal your time.
It's an art form, and it's about exploring the people behind the ideas as well as the ideas themselves.
I used to love reading the Saturday and Sunday magazines that came with the paper, this was back in the 1990s. Many of those were always of this long form, rambling, structure.
If you can see the pattern it's obvious once you start reading.
There's a sort of teaser at the beginning. Then they dive into a person's history like this:
Super Bionics is an exciting new form of prosthetic that's revolutionising lives.
Steve Jones was sitting on his porch when he first thought of super-bionics. Steve had always been fascinated by robotics when he was growing up. At 3 years old.........
5 paragraphs later they'll finally do a bit more about the super-bionics.
Then each section gradually moves the story forward in exactly the same way. They intro the subject. They introduce a new person behind the subject and explore them and their motivations. Then they say a bit more about the actual subject.
You're mistaking a writing style with time wasting.
I don't have the patience for it any more, but lots of people do.
Some of these articles are definitely more about the author being able to say they write than it is shedding light on anything or providing any kind of insight. It's all just gross to me
I don't understand why more people don't get this. I've told everyone who will listen in my org that implementing LLM's isn't going to solve the problem of people wasting our time reaching out to us with questions that have already been answered in our KA system. If someone was going to type something into an LLM, they could have typed it into a search bar. People don't skip the documentation because they can't find it; they skip because they don't want to read! They want to bug a live a person (and make it their problem)!
I was correct. Now we have a costly LLM implementation and have our time wasted with questions that are already answered.
None of the KB systems I've ever encountered, from orgs big and small, reputable or otherwise, had in them any useful information whatsoever. It's all just bullshit, roundabout sales, and a lot of answers to questions no actual human being would ever ask.
Whether or not to put LLM in the KB's search bar is immaterial. Myself, when I see a KA syste, I immediately close the tab and "bounce" back to the search engine.
> People don't skip the documentation because they can't find it; they skip because they don't want to read!
Yes, but also because they don't trust you to provide enough information to diagnose and solve their problem. There's no point wasting the time when experience tells you most customer-facing KBs are nothing but false hope and misleading headlines.
> They want to bug a live a person (and make it their problem)!*
Yes, but not because they hate that random support person - it's because it's the only remotely reliable way to solve the problem itself. Companies shouldn't complain, not after standardizing knowledgebases and phone menus, which are all implicitly and often explicitly designed to keep customers away from support staff - and instead of providing a solution, they're optimized to make the user think the solution exists somewhere and they're just too dumb to find it (therefore, user's own fault, not company's fault; customer's brand perception unaffected).
As for making it the support person's problem - that's literally their job. That's what they're paid for.
Maybe long form content solved a need back in the day when things were printed on paper and figured out well in advance, crossing their fingers on the relevance, and with where we are now we can suss it out without all the reading-as-middleman-to-knowledge
"Reading" an article through its comments makes the assumption that those commenting actually read and understood the article. This seems like a risk though, as there is an entire ecosystem of people who are just knowledgeable enough to be listened to by those with the same or slightly less knowledge of the content or field.
How many times have you sent a meme or made a referential comment about some piece of media that you've never even seen? Big Lebowski, Breaking Bad, and American Psycho memes are completely intelligible across the internet even though many people have never actually seen them.
I think the argument of the person that you're responding to is that these dilettantes would exist regardless of the tools that were out there, LLMs or otherwise. There have always been people that prefer to talk about things than to read and consume them.
The assumption that long form content is a relic and that reading is no longer necessary for knowledge seems absolutely crazy to me, but it does seem to be a common enough mindset that I've run into it with students that I mentor. It seems logical to me that if you could learn something in one hour, then by definition your knowledge in that subject can not be deep. But it seems like there are plenty of people that I work with and talk to that think a crash course or podcast is all you need to be an expert in something.
A rush to “get to the point” when dealing with art feels very much like the tech-obsessed productivity porn that can miss the forest for the trees.
A passage from E. M. Forster's "The Machine Stops" springs immediately to mind.
"... 'Beware of first-hand ideas!' exclaimed one of the most advanced of them. 'First-hand ideas do not really exist. They are but the physical impressions produced by love and fear, and on this gross foundation who could erect a philosophy? Let your ideas be second-hand, and if possible tenth-hand, for then they will be far removed from that disturbing element — direct observation. Do not learn anything about this subject of mine — the French Revolution. Learn instead what I think that Enicharmon thought Urizen thought Gutch thought Ho-Yung thought Chi-Bo-Sing thought Lafcadio Hearn thought Carlyle thought Mirabeau said about the French Revolution. Through the medium of these ten great minds, the blood that was shed at Paris and the windows that were broken at Versailles will be clarified to an idea which you may employ most profitably in your daily lives. But be sure that the intermediates are many and varied, for in history one authority exists to counteract another. Urizen must counteract the scepticism of Ho-Yung and Enicharmon, I must myself counteract the impetuosity of Gutch. You who listen to me are in a better position to judge about the French Revolution than I am. Your descendants will be even in a better position than you, for they will learn what you think I think, and yet another intermediate will be added to the chain. And in time' — his voice rose — 'there will come a generation that had got beyond facts, beyond impressions, a generation absolutely colourless, a generation seraphically free From taint of personality, which will see the French Revolution not as it happened, nor as they would like it to have happened, but as it would have happened, had it taken place in the days of the Machine.'
Tremendous applause greeted this lecture, which did but voice a feeling already latent in the minds of men ...
"Or perhaps what Terry Pratchett wrote about the river Ankh may apply: "Any water that had passed through so many kidneys, they reasoned, had to be very pure indeed." One has to wonder if people are thinking "Any idea that has passed through so many layers of minds has to be thoroughly refined indeed."
It’s like trying to explain what one may see hear or feel when their on vacation at an exotic new location by talking about the train tracks that brought you there.
So when you’re reading you’re not downloading packets that add up to some kind of point. Instead, in the absolute best case scenario, you’re simulating the experience, according to the author’s recommended doses, of someone else “acquiring” knowledge. This “someone else” is the nameless reader the book was written for but they are not you.
This seems extremely detached from reality. Just to give you an anecdotal example, I used to love reading books as a kid, but you'd be extremely hard pressed to find me reading one now. Clearly my reading habits have changed, and so it cannot be some intrinsic property written in my fate (which seems to be assumed to exist in the quoted framing).
Conversely, book purchases wouldn't fall or rise, especially over long timeframes, if there weren't changes in reading habits. It just doesn't make sense to portray people as these immovable objects, whose desire to read has been inscribed into them at birth.
> Summarizers start with the default assumption that reading is an obstacle standing between the reader and some kind of reward.
When reading novels for pleasure is not among your hobbies, it most likely will be. You bet I don't often want to read e.g. documentation: I just want to get my hands on the magic incantations or magic phrases / values required, and move on with my day.
But this is true even for literature class, with its mandatory readings. I was one of the more naive folks in my class, so I'd make an earnest attempt at reading through those things. At the end, that only mattered to the extent that I can now tell you it was a complete chore. And the "reward" at the end of those was good test scores of course - something others could replicate by just relying on, then manually written, summaries.
> one might be left with the false impression of understanding where there is none.
It is at most only an opinion that there is no understanding there, and so is that the impression the person would have about their understanding is false. Understanding is not a binary property, despite what e.g. the anti-AI crowd would make one believe, and you're no mind reader. I think that's pretty agreeable at least.
I do think your point that some people wouldn't read regardless is bad. People will waste time doing things that are available to them. If less forms of entertainment are available, picking up a book becomes more commonplace. I often wonder how much humanity's rate of progress will slow (or regress) in the coming decades.
- it's possible to transfer knowledge, as demonstrated by the fact that human civilization exists. It's not always easy, doesn't always succeed, and reading is a part of that, but it's possible and happening. I'm confused about your intended meaning in claiming otherwise.
- It's very difficult to distinguish between (especially one's own) understanding and a false impression thereof. To an overwhelming degree, the main realistic way is applying the knowledge, which is easiest when far removed from the activity of reading.
- One's upbringing, environment, social circle, etc., strongly influence one's propensity for reading, both for work and for pleasure. People change, especially as long as they're young, but even adults do in a major way according to conditions.
IIRC (I don't recall where I read about this), there are two problems:
1. "transfer" gives the impression that a person can copy their knowledge to another person, but that is not the case. The teacher says, writes the words or even demonstrates, but the brain in the student is making its own connections and tries to explain it in its own frame of reference. It may click, or may not, or may even click in the wrong way, leading to learning a different lesson from the one being taught.
2. The teacher may have tacit knowledge they do not know they have to teach, or convey by some other means. Most teachers don't even realize that this tacit knowledge is not present in their students.
So, maybe nitpicking a bit, but "transfer" is not the right word for it.
Before most people could read, you would learn a trade from your father or as an apprentice. Knowledge was handed down but you pretty much learned "the way it has always been done" and improvements were slow.
Once we all could learn from books and publish our discoveries, the spread of knowledge and the pace of advancement exploded. We went from farming with animal labor to walking on the moon in under a century.
These 25 screens of text are not for you, they are for Google.
Rhyming ending.
It's bad for the individual, but even worse for the collective. The AI summary reader isn't just convinced they understand, they also share that incorrect understanding in discussion. They effectively inject LLM slop into real life conversations and forget the subjectivity of reading.
This leads to what I consider more harmful. Discussions where the particiants themselves don't even believe in the stuff they are arguing. Where human beings, devoid of their own subjectivity, sling summaries and empty "facts" at each other. As if what 1984 textually said is important in any way beyond how you and I, the humans, connected with it.
i still like reading the adapted books into chapters from the walrus too. i havent been a subscriber in a while, but they tend to be nice reads that dont require me to commit to reading the same thing for the next couple weeks on the toilet
I had to learn the hard way 15 years ago that the average American adult cannot parse a full-page email in standard English. It seemed crazy to me at the time and seems crazy to me now, given that the average adult has completed elementary school, but most people are barely functionally literate at all.
I don't expect you to believe me. It's a weird claim. But walk into any average grocery store and hand someone a page out of a book and ask them to read it out loud to you. Many people are so aware they can't that they will refuse to try. Of the ones that do, you will struggle to find one who can read the text with anything like the fluidity or inflection they would use to speak the same words. If you give them time to prepare, they'll probably be able to get through it in a few minutes, but nobody's putting that kind of effort into a text-only email, even if it's important for work.
Reading is so difficult as to be a chore for the average person. They don't just see written words and know what they say. They really have to work to get meaning out of written text.
With the proliferation of other means of taking in information, many of which require no effort of any kind beyond hitting play and staying within earshot, why would people choose to read? They didn't want to do it before. And now they don't need to do it either.
Of course.
I would likely find this situation very unsettling, if not stressful. I would probably be caught off guard. I would probably perform badly if I accept at all for all sorts of reasons unrelated to my ability to read smoothly. Unless I'm feeling particularly positive and gather the energy to pace my mind, apply myself, enter a character like I'm having a role in a play, and remember to be slow, and to forget about the content. Reading out loud is more complicated than simply reading for yourself: you are, at the same time, both reading and articulating speech.
What's more, when you speak, you are using your own words, your own (oral) phrasing, and you know where you are going, or at least enough to have a decent prosody. When you read out loud, you are reading phrasing from someone else, which may make the prosody less smooth. And if you haven't read the whole sentence in advance, or at least ahead enough, you may struggle.
Independently of knowing how to read (fast), reading out loud probably takes practice to be smooth.
With your experiment, you are testing all sort of things unrelated to knowing how to read, to the point that you can't draw any conclusion on the ability of people to read.
Parsing a text also depends on the writer's ability to write well. If the text is boring or its phrasing is overly complicated, yeah, it will be difficult to read.
We also live in a world plagued by focus disruptions. You are not only dealing with people's ability to read, but also their ability to remain focused… in a setting where they possibly get interrupted very frequently.
For all sorts of reasons, reading a wall of text is indeed hard. This includes the reader's environment, the presence of their smartphone next to them, how tired they are, whether they are concerned by something else, whether the text is actually interesting to them, and how well the author of the text writes.
Even "caught off guard", not "feeling particularly positive", and "reading phrasing from someone else", some people have so fundamentally mastered reading as a skill that this wouldn't be difficult for them even if the challenge were attempted drunk. The point of the parent comment was, I believe, that for such people it may be hard to imagine that others would believe "reading a wall of text is indeed hard" because it isn't for them.
As an experiment I spoke your comment aloud as I read it. It sounded natural and I made no mistakes at all.
Deleted Comment
Reading is a skill, and it's not the same as being (technically) literate. For someone who's a competent reader, there's nothing at all more intimidating about being asked to read a page from a book (you can choose the book, but a page from a book is not a "wall of text," and the book's publication also implies at least some pre-screening for accessibility) than being asked to have an impromptu conversation about where to find the potatoes, and whether you know any tasty recipes for them.
Most people don't read well. They read well enough to understand street signs before they pass them, or to get through an article intentionally written to a 6th grade level. But anything where the details are important? Or that requires explanation? They either cannot do it, or find it so difficult that they won't.
How is that relevant ? Reading out loud is not the same as reading ? I struggle to keep pace with reading out loud - because I am just not used to it. I do better now that I read to my kids, but I don't think I have reading problems - reading out loud is a skill that's on top of reading. Speech patterns, words you're used to pronouncing, etc. all make reading out loud more difficult than speaking or just reading IMO
The underlying truth here is worse than 'majority of educated are illiterate'. Collectively, we've built these delusions into our culture. Perhaps there is less suffering this way.
I knew "most" American adults couldn't. I didn't realize it was 87%.
Is this tru’ish? I’m not refuting it I’m just a little shocked this might be the situation we’re in. I know generally people now struggle to consume long form content but it even being able to read a story?
That stranger might conclude I can't read, when in reality I devour books, am the beta reader for my two published-author friends, and probably edit the "thoughts on books I've read" section of my personal wiki more than any other.
> I know generally people now struggle to consume long form content but it even being able to read a story?
There is reading and then there is reading an unknown text out loud in public while being judged. The two are not the same thing. I had to read publicly when I was a student years ago when smartphones were a new thing. They handed me the text with instructions to reread it at least twice ideally out loud before performing.
The point is, it is not like 20 years ago a good student would be expected to read super fluidly out loud in public without at least a little preparation.
Deleted Comment
> I had to learn the hard way 15 years ago that the average American adult cannot parse a full-page email in standard English. It seemed crazy to me at the time and seems crazy to me now, given that the average adult has completed elementary school, but most people are barely functionally literate at all.
They can, they just won't, because they don't give a fucking shit. The moment you hit adulthood in modern times you're bombarded with bazillion of bullshit. Do you seriously believe your meticulously hand-crafted email is high enough on someones radar that they'll actually pay attention?
I don't know that what you've described is any different now than 20 years ago when I was in high school. People struggled to read aloud texts like plays or classic literature. I would use that as a bar for complex prose that benefits from good narration; often the point was to encounter unfamiliar words or meter and to read with the purpose of critical analysis. A friend of mine is an author and we did a group reading of a play they'd written. Quite hard to do off the bat. Similarly if you've ever tried to DM a role-playing game like DnD, where the text you read is semi-randomly chosen based on what the characters decided to do (also, will you role-play dialog?).
I've worked in academia for 10+ years at this point. You can tell almost immediately if a presenter has had practice, knows their material or is comfortable speaking in public. Lecturers and professors are, unsurprisingly, often very comfortable giving presentations and there are people who live for conferences and working groups. We're required to read dense material frequently. Understanding of a piece of work, or the attention span required to ingest a scientific paper, does not necessarily mean you could read it aloud fluently.
It's obvious if you think about learning a foreign language. Parsing written text silently is a completely different task than producing speech with the right intonations, even if you recognize the words.
Reading aloud fluidly requires a certain speed and anticipation. You have to be reading the next word in your head while you are still pronouncing the previous one. This isn't necessarily correlated with reading comprehension which is purely input.
> But walk into any average grocery store and hand someone a page out of a book and ask them to read it out loud to you.
I doubt that you have done this, but if you have please stop doing it. This is literal insane behaviour.
With Claude as I read I can constantly check my understanding. When my response elicits a "Well, not exactly..." I know I have to go back. This combined with the ability to have Claude clarify formula details from a phone picture has rapidly accelerated my learning and has me reading much more these days.
Claude is also pretty good at subject specific recommendations, especially when you're looking for a specific type of treatment of a subject.
The authority received the large body of text but, due to time commitments and attention, they didn’t have time to read it all. They used AI to convert the text to a concise bulleted list.
Not the point they wanted to make, but a point nonetheless.
Deleted Comment
Haven't seen so many pitches for summarizers lately.
I haven't used chatgpt (or whatever) for summaries in a couple of years, so i have no idea what SOTA is; although "chat with a document" seems like it'd be more useful in general than a summary for the way i eschew long-form articles.
I traced down this through an academic article which favourably compared it to other summarising solutions back in 2006. It might help answer the question: https://web.archive.org/web/20070209101837/http://www.copern...