AI Is Already Killing Books | Hacker News Reader

AI Is Already Killing Books

matduggan.com

85 pointstschumacher155 comments

This was a problem before "AI" and will definitely be a bigger problem going forward. That said, there are more great books than I could read in 100 life times published before 2020, so I doubt this will be an issue for me personally in practice. For current events or evolving knowledge like new science or technology, I would probably have that summarized by an LLM anyway as a preference. The author is really saying "AI is hurting authors".

Your post makes me sad.

"The Value Of Owning More Books Than You’ll Ever Read" - https://clivethompson.medium.com/the-value-of-owning-more-bo...

https://archive.ph/G7H9s

"...The writer Umberto Eco belongs to that small class of scholars who are encyclopedic, insightful, and nondull. He is the owner of a large personal library (containing thirty thousand books), and separates visitors into two categories: those who react with “Wow! Signore professore dottore Eco, what a library you have! How many of these books have you read?”* and the others — a very small minority — who get the point that a private library is not an ego-boosting appendage but a research tool. Read books are far less valuable than unread ones. The library should contain as much of what you do not know as your financial means, mortgage rates, and the currently tight real-estate market allows you to put there..."

This was partially solved by the internet and Google Books although there's still a lot of room for improvement. We really need to get all of the books in the world digitized and behind an API that allows anyone to access all of the content with some sane pricing scheme rather than charging per book. I doubt this will ever happen so the best we can hope for are the various efforts to pirate as many books as possible and make them available for download. It would be a dream to have a really good search over all of the written works of mankind though.

Yes, but some books require extensive effort to write while others are produced factory like. How do we take that in account?

It shouldn't be, the effort with which something is produced has zero weight when judging its value. For books they should be rated based on their purpose. Non-fiction should be rated along dimensions of accuracy, new information, utility, clarity and absence of errors. Fiction should be rated on how entertaining and inventive it is among other things. If you want to subdivide further I'm sure you could come up with other meaningful criteria.

> This was a problem before "AI"

Yes, I knew someone who was in "publish as remainder", that is publish at a high price then sell in bulk at a discount. Creating the books involved:

1. Think of a subject

2. Get some pictures

3. Find someone to write something which goes with the pictures

4. Repeat

There has always been a market for the undescerning.

Contemporary literature is valuable for its insights in the present. I love classics, and much of what I read was published before this century, but I would be missing out if I stuck just with tried and proven titles and authors.

I would say that contemporary literature can't provide any insights, by definition. Because it's a slave to its zeitgeist. Anything touching politics, history, economy and especially society will be tainted.

Wouldn’t this apply to all literature then? Because it was all a product of its time.

Of course all literature is a product of its time but when you are reading an old book most of the time you are distanced from the events or then-new ideas presented and it's much easier to think objectively about them. As an example I'll give Marx works on communism. If I was XIX century working man I would be ecstatic about the idea, with the hindsight how communism actually works in practice I would have second thoughts. But the difference between the theory and the practice would be the actual insight. And in case of the contemporary events the actual insight would be in how they were presented and perceived when they were happening compared to how they are seen now

Marxism is very antiquated without modern insights applied to it.

Reminds me of The Machine Stops:

    “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. [...] And in time there will come a generation that had got beyond facts, beyond impressions, a generation absolutely colourless, a generation seraphically free from taint of personality.”

The difference is that the books we still read are the ones that have proven to be more universally true than the rest. We only know about old books that have remained relevant in one way or another. We've long forgotten the tripe.

With modern things, we don't yet know what will remain relevant and what will fade into obscurity or lose its relevance.

Plenty of still-popular ancient books are full of fiction, myth, and superstition. Perhaps one could say some have remained more useful.

That isn’t an argument to stop producing them. It’s an argument for letting time pass to determine which remain relevant.

Can you give an example of a book published after, say, 2000 that is really valuable for insights we would not be able to get from either older books (if related to the human condition) or news articles (if more to do with some bare fact)?

I read mostly books written before the World Wars and I’m doubtful there’s much after that period of any real and lasting value. That doesn’t mean I don’t enjoy some of it, of course, mostly the fiction. The exception might be really niche works of local history, which is probably <0.001% of all such books, or a few really good scientific/mathematical compendiums.

Obviously it's a bit of a cop out but analysis of events. Books on the invasion of Iraq, or the gulf war. Books on the financial crisis. You might say this is a small portion of what's worth reading, but these are things with undoubtedly huge relevance to what is happening today in the short term future, as well as giving insight into the people and systems that govern us very specifically

I could be wrong, but I think almost everything in those examples exists in news articles, although to be fair, those articles are often long-form, and some are sourced with reporting from the books you mention. I've never read a book on either subject, but I maybe presumptuously think myself well-informed. ;)

I do think, as time goes on, and things are declassified, we'll learn more about e.g., the Iraq War than we know now. To a point, things become clearer with distance, which often renders books written on contemporary subjects outdated. But a book published today could be a decent summary of events for people who didn't live through them, or were too young to pay attention at the time. Unfortunately, one recent why I don't pay them too much mind is that most of the genre is hopelessly partisan, one way or another, and I don't know that there are any really great historians to write something timeless. But either way, I would say the insights you can glean from those books are still readily available if you're watching events critically today. Not much has changed, and if it has, it was usually for the worst. Maybe the most interesting part for many people would be the reminder that many of the same people directly responsible for those disasters are still considered respectable people today and help drive policy. E.g., Bill Kristol didn't go into some kind of exile - he's still considered a very serious person and has a lot of influence and is out there shilling the same type of thinking in a different context.

> I could be wrong, but I think almost everything in those examples exists in news articles, although to be fair, those articles are often long-form, and some are sourced with reporting from the books you mention. I've never read a book on either subject, but I maybe presumptuously think myself well-informed. ;)

Crashed, by Adam Tooze was a very very good retelling of the GFC, and I did read all the newspapers at the time and still found lots of interesting stuff.

More generally, I'd argue that modern historians probably have the highest chance of producing useful new insights recently.

If you accept fiction (and like fantasy fiction), I'd argue that the Malazan books of the fallen and/or the Nine Worlds series are serious works that are going to endure. Ask me again in a century, though ;)

I don't think it's as bad as <0.001%. There's a larger volume of books published these days than ever before the 2nd world war, so there's a lot more crap to wade through, the nice thing about old books is that time has done the job of pruning out a lot of the crap.

I think people still write good books which can provide good insights into the human condition, it's just hard to find them in the present.

It's not that any specific particular work of fiction will provide really valuable insights — what is insightful to one reader may be obvious to another — but that the whole field changes and absorbs societal change and extrapolates and interprets society from there.

It means books get written now which explore perspectives and voices unheard before, which in turn can help us readers expand our frame of reference.

The collective works of fiction written before 1900 tend to reflect the societal viewpoints of well-off white men (even when written by women or specifically dealing with societal ills). Go a few decades beyond that and you see authors from a working class background join the chorus, then more women, a broadening of sexual themes reflecting society's change (feminism, sexual liberation, homosexuality, etc.), more open criticism of religion too. Digital technology changed society significantly, and this is of course reflected in writing from the more recent decades, and coming up towards today you see more and more diversity amongst authors, adding — through the characters and narratives they create — yet more perspectives and insights. Sometimes pushing the envelope of a specific field, sometimes getting rid of tropes which no longer convince. Fiction changes constantly and will always be rooted in the year it was written.

You sell yourself short if you stop at 1940.

Unique perspectives are not made by sex or skin color, but by life experiences.

Two Americans today, whatever their sex and race, have virtually everything in common with each other compared to anybody from 1800, let alone 800. The extreme, excessive focus on race and sex in contemporary writing is exactly what makes it boring and irrelevant. This comment is a great example - you've been taught by contemporary writing that such a tendency existed, when it in fact excludes the objective reality of the tons of works that could not have been said to be "written" or "about" such people even by modern framing, but also the fact that "well-off white man" is a completely meaningless and inapplicable phrase if you go back more than a couple centuries.

The sort of work you're describing is the stuff we're taught to acculturate us to the world we already live in. There's no point browbeating me with even more material that I am already steeped in.

Sex and skin color most definitely influence unique experience (not exclusively of course). I guess you didn’t catch this in your readings of Shakespeare.

> Two Americans today, whatever their sex and race, have virtually everything in common with each other compared to anybody from 1800, let alone 800.

Define "virtually everything"?

The rich, white, and men have had massively different resources and societal privileges in each of those eras.

Even today it seems obvious to me that to be rich or white or male each brings benefits at every stage of life which can drastically change ones life experience: food security, personal safety, education, job prospects, and romantic opportunities.

I mean that if you put a random 2024 white man and black woman in a room with Charlemagne and the ability to communicate, the former two would have 99% overlap in their worldviews and perspectives while Charlemagne would be an alien to them both. “Privilege”, as the ancients noted, is fleeting and superficial.

“White” is a modern social abstraction that rapidly breaks down the further back you try to apply it. It also carries with it loaded assumptions that do not really hold even at a population level in different places. This is exactly what I mean: immersing yourself in modern culture blinds you to the reality people did not and could not apply such categories because they did not even exist. People often repeat that race is a social construct but rarely think about what that really means.

People in past cultures were totally alien. There’s some substrate of common humanity but trying to say, apply modern racial privilege politics to Xenophon’s encounter with the very pale Paphlagonians, or the widespread trafficking of European slaves even into modernity in Northern Africa and Anatolia, or even into parts of colonial America is just nonsensical.

I mean, take this quote from Benjamin Franklin for example, which would have been early in the conception of modern race:

> Which leads me to add one Remark: That the Number of purely white People in the World is proportionably very small. All Africa is black or tawny. Asia chiefly tawny. America (exclusive of the new Comers) wholly so. And in Europe, the Spaniards, Italians, French, Russians and Swedes, are generally of what we call a swarthy Complexion; as are the Germans also, the Saxons only excepted, who with the English, make the principal Body of White People on the Face of the Earth.

This is a completely bizarre Martian take to any of us alive today. The past was a foreign country. If you just stamp your feet and repeat “all old books is just rich white men” you’re not only completely wrong but missing genuinely different perspectives.

As an aside, what does it mean to not have a reasonably good understanding, as a man, of a woman’s perspective in 2024? Did you have mother, sisters, or cousins? No friends or girlfriend or wife? No empathy? As a last resort never read mumsnet or 2X? Ignoring the fact that once again highlighting the sex differences ignores the fact there are literally four billion women and they all have different perspectives, they’re not a hive mind, if you have to read a book to figure out the common differences in perspective…hasn’t something really gone wrong in society for that to even be possible? And could a book by some woman you don’t know and had some kind of connections or very unique experience - she got published after all - really tell you more than having long and deep empathetic conversations with people you actually know?

Books are great because they’re a window into that we can never see for ourselves. Literally 50% of the people on Earth right now are a different sex. You can find out what they think right now with no effort. We don’t need the nth book on “what it’s like to be an X in America”, we all know, everyone is shouting it from the rooftops. I’d trade it all for more Sappho, and I’m a member of a minority group that could write my own What It’s Like to Be An X in America book.

I recently read a blog that argued exactly the opposite: that there was no point in reading old books because if they contained good ideas, somebody had written a newer book that better summarized those and combined them with new ideas, and that old books were liable to be wrong. Both arguments are wrong. The point is to read good books, not (usually) to just pick books before or after a certain date. I see elsewhere you argue that people at the time can’t evaluate the facts because they don’t have perspective. It is usually argued that people writing history lack understanding (and facts) about periods they didn’t live through. Short vs long form is also missing the point: short articles can be error ridden, misguided and even boring. At best they will waste less of your time, but there are great, long books that provide texture and insight that cannot be summarized.

The proposition that no one has written a worthwhile book or had a new insight in the past 24 years is so laughably indefensible that it stuns me to see someone put it forth in earnest. To engage with it risks dignifying it beyond its stature. Rather than post my own list of books, I will link to this democratically determined list of "books that changed your thinking, changed your mind, enlightened you, helped you move ahead, helped you heal, in short, made a difference.":

https://www.goodreads.com/list/show/21995.Best_21st_Century_...

I don’t guess we’ll ever come to an understanding then, because I find that list totally laughable and a perfect example of what I’m talking about. I like Bill Bryson and I own the book but if his a Short History of Nearly Everything (#1 on the list) is the strongest example of what “changed your thinking, changed your mind, enlightened you, helped you move ahead” etc., well, things are very grim indeed. Scrolling down the list mostly gets much worse. The world would be no worse off if most of those books had never existed, and many of them don’t offer anything new at all. I scrolled down to 100, against my better instincts and encountering such luminaries as Tina Fey.

People have had interesting insights in that period - of course - but very few of them require anything longer than a sentence, and they are few and far between.

By the way, 2000 was chosen semi-arbitrarily, I would probably put the line much farther back.

This is a general problem with art, music, movies, books, etc. There is just so much!

Performances and storytelling was pretty much ephemeral before printing presses, records, and now computers. Now it’s cumulative. All the writing and art of the past several hundred years is out there.

Kind of creates an overproduction / saturation problem.

Zoe Bee did a video essay[0] on her experience as a ghostwriter producing essentially the kind of junk that is now being automated with LLMs. In her case she was (somewhat unknowingly) paid to write unlicensed Minecraft stories. I think the video is worth a watch if you're not aware of the state of ebooks (especially on Amazon) prior to LLMs.

It's also worth mentioning that this problem isn't limited to ebooks. There's also a cottage industry of mass produced minimal-effort audiobooks on Audible as part of various "passive income" scams. Dan Olson made a pretty good video essay[1] about one such scam where he actually played along for most of it and also gave it a try as a ghostwriter.

[0]: https://www.youtube.com/watch?v=O1aqLLiIjgA

[1]: https://www.youtube.com/watch?v=biYciU1uiUw

EDIT: Considering the most cited examples are Amazon/Kindle and Audible, I think parallels can also be drawn to the proliferation of no-name brand Chinese whitelabel dropshipping products on regular Amazon. Everyone already knows not to trust Amazon reviews but at this point it's hard to find reputable brands for products you're not already familiar with (e.g. which of the one hundred brands featuring near identical products are actually brands you might find in a retail store rather than a random name slapped on the product in the same Chinese factory?). LLMs will definitely make reviews even more untrustworthy but they might also help generating even more plausible copycat product descriptions and designs.

> e.g. which of the one hundred brands featuring near identical products are actually brands you might find in a retail store rather than a random name slapped on the product in the same Chinese factory?

Ye this one is annoying. I often look for automotive tools and before I realized this I thought I was turning insane. Different store fronts pretend they are making some tool or like has sourced a factory to do their design.

But they like order the tool with a sticker and paint job from the same supplier.

Sometimes nuts and bolts or like cover plates vary but it is the same base.

I was browsing for tool hooks for the garage the other day and it felt like there are two factories in China that make hooks and two factories that put rubber on them for 4 combinations and that is it. But 20 flavours of branding.

Edit: It would be really neat if the factory had to mark everything they made.

I recently came across the tongue-in-cheek conspiracy theory that consumerism has reached its endstate in that companies now intentionally design only good-enough-but-unsatisfying products because those make you more likely to continue shop around for something different-but-equally-unsatisfying, so just like planned obsolescence but more based on dissatisfaction.

I think there is some truth to this in that it's likely profitable to flood the market with mediocre products (especially if you do it through a white label network of "competing" brands) as long as you can avoid refunds (and on Amazon certain sellers make returns intentionally difficult by providing Chinese addresses without a prepaid label, making you pay more for the return shipping than you likely paid for the product to begin with). This also drives down the overall expectation of quality and reduces market pressures for quality while also completely swamping any competitor trying to sell a genuinely high quality product by making reviews completely unreliable.

I don't think this is a coordinated strategy as with whitelabel dropshipping being sold as a get-rich-quick-scheme for years there's no real need for it. Plus as you mention in many cases this is simply a consequence of a race to the bottom earlier in the supply chain resulting in nearly identical mediocre products even when some parts differ.

Personally I've run into this when picking out interior doors for our new house: despite being in the mid-tier price range, various aspects ranging from the veneer to the locks are extremely underwhelming but apparently on par with other doors in that price range. There's literally no reason the locks should feel like cheap plastic toys but it's something that's not visible and can't usually be tested even in a show room so apparently that's where they decide to cut corners. And because even standard sizes are considered custom, there's no way to return them, let alone once installed.

Ye. It is hard to give a number of significance of the effect, but it is interesting.

> I think there is some truth to this in that it's likely profitable to flood the market with mediocre products

I think a factor here could be that you don't want the customers to recognize bad products? Some appliance models rotate so fast there are hardly any (non fake) review consensus on them. It is not just no-name Alibaba, but "proper" brands to, that seem to do that.

Current LLMs can not summarize books even if they are in their training set. It is only possible if the training set also includes human summaries.

> Current LLMs can not summarize books even if they are in their training set

Why is this? due to legal reasons?

The context windows aren’t large enough, as I understand it. It might be possible via a chain-of-summarization, though.

Most importantly due to the context length not being long enough. If the context length was long enough, it is possible that they could do it with clever training. I only trained much smaller language models though.

"Even if" is an odd phrasing. Why would you expect that to be possible? LLMs don't have direct access to their training set. Having a book in the training set just means they emit text in the style of the book - it's not like a database they can query. But they can summarize very well, if the text is front of them.

I mean, currently you have to wait for a human to summarize a book, before you can read an llm rephrasing of that summary. Or you have to buy the book and pass it to the llm employing some tricks to avoid reunning out of context length.

The complaint that too many books are being written and that most of them are useless nonsense is indeed as old as time -- or at least, as old as books.

Montaigne (1533-1592) famously said il y devroit avoir quelque coerction des loix contre les escrivains ineptes et inutiles ("there should be laws against stupid and useless writers").

This was about just a hundred years after the invention of the printing press.

I remember wondering what books on the stock market would have been like before like the 50s as a curiosity. I've taken 2 CFA exams and have a Finance degree and noticed that none of the major concepts introduced seem to go back further than the 70s and market was definitely older than that.

So I searched on Google Books and limited the timeline... What I found were the pretentious ramblings of an old man that thought way too highly of himself for getting lucky at a time that a little luck could make you a fortune to last a lifetime. He wrote a bunch of books on the market. It was like 1910s or something. They were so worthless, but I did find them hilarious to read.

Here's a couple of well known ones. Not sure if they were what you found:

https://en.wikipedia.org/wiki/Reminiscences_of_a_Stock_Opera...

https://en.wikipedia.org/wiki/The_Intelligent_Investor

Nah, those are different. Especially the second one. lmao. It also happened to have few sales till a bit later on. I think my goal back then was to see what someone would find if they were just trying to learn about it at the library. Reason why is because the ROI using the Fama-French data back then would suggest someone could have rapidly become a millionaire and billionaire off of a thousand dollars if they performed even the most basic value investing formula back then to bifurcate the market or, better yet, divide it into deciles based on (book equity)/(market equity). Only in the past 10 years, you'd actually underperform the market return.

So, considering a very simple formula could have been so effective, I wondered what they'd find at the library and if even that simple concept would have been available. What wouldn't be available is SEC data. But, as your second suggestion made clear, you could get the information for ratings agencies.

I mean, comparatively speaking you were looking for programming books in an era where hardly anyone even knew what a computer was. The market changed so drastically at the turn of the century that it really only shaped into our current recognizable form in the 50s and 60s. There's many reasons for that.

To be clear, I expected NOT to find anything that would help accomplish that task. Because the vast opportunities would imply it didn't exist. So, yeah, what I expected to find is crap and what I found was hilarious crap. The dude did have a way with words, though. It's just hilarious how pretentious the books were considering they didn't really impart any useful knowledge.

GenAI is not making a qualitative change in spam & junk content, it's making a quantitative change. Previously, you had to wade through some noise to get to the signal, but with everyone making scammy content with genAI you'll have to wade through 100x noise to find the 1x signal.

Just like existence of email didn't create the concept of spam, it just made sending it much, much, cheaper.

> That said, there are more great books than I could read in 100 life times published before 2020, so I doubt this will be an issue for me personally in practice

I spam this fact whenever someone makes arguments against piracy. I am utterly fine if no new culture is produced, as there already exists more than I could ever enjoy. Besides, creatives will create (with lowered production values) regardless of money. It’s a primal drive for them.

> I am utterly fine if no new culture is produced

Although, I would be a little upset if they stopped making high-budget binge-worthy series.

Then again... What the hell do I care if it's made by AI? Especially if they use one of those preference-based RLHF and penalize the model against episodes where the series lost all its viewers.

The same can be said for most code. It isn’t improving quality of life for most people. Like I really need another todo app. Let all the programmers live on scraps. It’s a primal drive for them. /s

This but without any irony.

I wonder at what point people would consciously spend less time online and go back to physical books, especially those works that were published decades/centuries ago? If I am a normal, non-tech person reading something online, how am I supposed to know who wrote it (software or human), what is the agenda behind it etc?

Of course I understand physical books can also be written by AI (just last week, I saw a physical poetry book fully written by AI, except for the preface). But it is much more cost to produce a physical book than throwing up something online only.

>That said, there are more great books than I could read in 100 life times published before 2020, so I doubt this will be an issue for me personally in practice.

I'm personally split on this viewpoint. Through a series of events this year (starting with the death of Charlie Munger), I've started reading the Harvard Classics. Doing the 15 minutes a day challenge, which started on New Year's Day. I bought the full original 1910 set on Ebay. They're pretty cheap because they're essentially just room decorum and probably nobody reads them. People on Ebay will gladly sell cool looking leather cover books because it makes their offices look nice. This is an actual descriptions many sellers use.

For me though, above all else it's been an incredibly humbling experience. Because its been a very long time since I've read works from genuine professionals who spent a lifetime mastering their craft. I'm struggling with prose and vocabulary. I'm not an avid reader. I will at most read maybe 5-10 books a month. Some of which are technical or industry related. But I still like hardcopies so I tend to visit bookstores. And the crushing reality is that what's generally available on bookshelves (at least in the USA) is fucking garbage. Half-assed ghostwritten autobiographies of no-name celebrities or politicians, lifestyle in-your-face FEMALE EMPOWERMENT books that give questionable advice, pseudo-historical nonfiction books, bland and uninspired Star Wars science fiction rip offs (and their accompanying video game expanded universe novels). The list goes on.

Barnes and Noble realized that their audience doesn't really buy physical books anymore (or just plainly don't fucking read) so have moved in to fill the gap that Toys R' Us left behind. Half the floor space is a glorified toy store. Most of the people that come here either bring their kids or are kids themselves hanging out in the Manga section.

The cliche is obviously here and repeating itself: but with each consecutive generation we become less literate. AI has made it so the illiterate get nifty summaries of other less literate summaries of great works and ideas, which then get challenged by the pseudo-intellectual hack frauds extracting wealth from their fanbase and college students pumping out bullshit papers to meet graduation requirements. Just how fucking low exactly can we limbo before it's too late? In this context AI is basically Accelerationism

Yeah, not too worried. AI might kill the "self-published Amazon eBook market of garbage genre novels", but honestly there's very little value there. The suggestion that an LLM could write a Hilary Mantel novel, or a Robert Caro biography, is so far beyond ludicrous that it's not even worth arguing.

I don't know who these people are that are reading books so bad that an LLM could plausibly replicate the work, but if that market is killed by a swarm of AI-written nonsense, honestly: who cares?

> Now that AI books exist, the probability that I will ever blind purchase another eBook on Amazon from an unknown author drops to zero.

This is a thing people do? Like... before AI? Not even reading extract or anything? You just see a cover and a title and go like "sure, I'll spend my hard-earned cash to make a 30 hour investment in this thing I know nothing about"?

The best version of this argument you can make is that it's about the treadmill: great writers aren't born great writers, they have to write a lot of crap first to become great writers, and this market is how you do that. Take that away, you don't get any more great writers. But I don't particularly buy that either: there are very few writers I love that were able to successfully make a living selling self-published Amazon eBook garbage until they got good enough to actually be picked up by a publisher.

Journalism, however, is a different story: plenty of great writers (fiction or non-fiction) got their starts as journalists and honed their craft writing small pieces there, and that is a market that is under total threat from AI. That's maybe a cause for concern. But I don't think it's an existential threat to literature as an art form. I don't think that's ever going to go away.

I can tell you who is reading books every night that are absolutely terrible and don’t require a coherent plot! Children. My kids and I have been crafting a few sentences of some absurd plot then reading a story about it generated by ChatGPT.

My daughter got so excited that she drew pictures of the Chicken Nugget and Cookie brothers and showed them to me after work, it’s like a choose your own adventure book but you can really actually choose your own adventure.

Personally I find the stories vapid and boring but before I was reading some brain dead story about Ariel using her favorite Dinglehopper to brush her hair, so whatever.

We are also reading Little House on the Prairie which I consider a little more “high brow”, and the girls really like that, too. But kids are easy to please and it has been interesting.

That's a fair point, children's literature (a wonderful and valuable genre) will certainly be under threat. I would probably suggest that nothing an LLM would come up would stay with a person for decades like (say) Were The Wild Things Are. I'm Swedish, so I grew up on Astrid Lindgren, and some of those books move me to this day when I think back on them. I imagine many Americans still find Dr. Seuss hilarious in the same way.

> We are also reading Little House on the Prairie which I consider a little more “high brow”, and the girls really like that, too. But kids are easy to please and it has been interesting.

Yeah, that's kind of my point :) do you think an AI could write Little House on the Prairie? I don't. Sooner or later the vapid and nonsensical plots from LLMs stop being that amusing, and you need to feel real human connections to the things you're reading. That's the reason you read books, after all.

An LLM seems by far the most likly to be able to write a Hilary Mantel novel to me. May need to be a fair bit biger then ChatGPT-4 but given no living human ever has it dosn't have much compitition.

> self-published Amazon eBook market of garbage genre novels

"Even after two months on their prolonged, glamorous, non-sleep shopping spree, she couldn't resist the allure of his white shirt over his hard pectorals. He wasn't at all like the other men in her life, vapid things that would spend a lot of time in the gym trying to lose weight, instead of just being outright gorgeous and indulging her social needs."

A part of me agrees with your statement: people indulges themselves a lot in junk literature, as they do in junk mates and unfulfilling social events. And yet, I have known myself for enjoying junk food and junk literature and sorely missing partners who were certified by their backwater witch-doctor from Hornborga as having too small of a brain for even a minor demon to consider possessing.

We may end up in a world where some people will enjoy AI-generated content, and some people will be at war with the death of the soul that comes with it. I'm decidedly in the second camp.

> Yeah, not too worried. AI might kill the "self-published Amazon eBook market of garbage genre novels", but honestly there's very little value there.

I'm not convinced. Speaking as someone who's ended up selecting for a bunch of the good (i.e. picked up by publishers) authors for self-published genre novels, I definitely feel less likely to actually put in the effort after a bunch of bad experiences, and certainly if it's LLM generated.

I'm pretty sure I'd nope out, but because there's so much crap it's hard to pay enough attention to find the gems.

At this point, I'm selecting based on prior knowledge and the Hugo longlists, which is depressing as it selects against the interesting new stuff that I'd like to find.

> This is a thing people do? Like... before AI? Not even reading extract or anything? You just see a cover and a title and go like "sure, I'll spend my hard-earned cash to make a 30 hour investment in this thing I know nothing about"?

Yes, many many times. Mind you, I tend to exclusively focus on books for entertainment, and most books aren't that expensive (especially after spending a few years buying academic published statistical books) but I agree with the OP in that I'm becoming less likely to do this (but to be fair, this was happening before LLMs became super convenient and popular).

Actually, thinking about it, I find it pretty unlikely that any current LLM could actually write a coherent novel, given the context window. You'd probably need to do some kind of chain to make it work, i.e. generate synopsis, then recursively generate more text. It would also likely be super inconsistent unless you kept feeding the previous parts in.

Hmmmmm....in a world where I have more free time, I might take a stab at getting LLaMa to do this.

The recent advances in LLMs have made me realize how important a signal well written English was to filtering content by quality. If the first few sentences were on topic, syntactically correct, and grammatically complex, it was worth skimming at least. This was obviously not remotely perfect, but it was pretty good. LLMs are blowing that up. They have not increased the amount of garbage on the internet (plenty of stupid people around to supply that), but they made it a lot easier to make garbage look superficially like it was thoughtfully written by an intelligent person.

That's not the first thing people might worry about, but this will also be a pain for archives.

When slop floods the library faster than it can expand, who will want to maintain that. I don't think we have good enough sorting and rating (or good enough AI output detection) to prevent bookspam.

This is a point in history where "record everything" stops being viable, and we have to start hand-picking the text we want to survive. Indiscriminate things like Internet Archives stop being viable.

Why keep books, anyways? You can just ask the AI to re-generate whatever it is you want to read about on the fly.

In France there is a law called "dépôt légal" that forces all publishers to send a copy of each new book to the national library for archiving. It's been in place since 1537.

But ebooks are exempt from it. (Ebooks obviously didn't exist in 1537 but they've now been around for decades.)

So unless AI is used to make books published via traditional publishers (which is of course possible, but somewhat unlikely), the situation doesn't change much as regards to archiving books (in France).

I'm not sure of the situation in France, but this is starting to happen readily on Amazon. Lots of physical books you can buy that are completely AI-generated nonsense.

Are those books from real publishers or from Amazon's Print on Demand service?

I should have been clearer; it's not just ebooks that are exempt, but anything without an ISBN. You can make a paper book with Amazon print-on-demand that doesn't have an ISBN, and then you don't need to send it to the dépôt-légal.

Many people argue this is bad and should be changed. But it's debatable.

According to https://www.bnf.fr/fr/depot-legal-pour-quels-documents, pure ebooks are not exempt; they are just archived directly from the distribution platforms by sampling.

What scares me is that some mixed works (PDF/on-demand prints) are not archived because the authors think the system is too clunky to deal with, which is really a shame for books like https://laurent.claessens-donadello.eu/pdf/lefrido.pdf, for example.

Of course we have a mechanism to filter books: publishers. They

Now ebooks and on-demand printing allow you to "publish" books without going through the quality filter of a publisher. But you won't get into a book store that way, and anybody talking about the trouble of archiving those better already have an archive of the low hanging fruit of the same category: fanfiction.org and friends (and I don't mean this derogatory, there are great stories in there)

Libraries are already flooded. Libraries regularly send older books to be pulped, due to the fact they often contain outdated perspectives. They also look at last-checkout. There's a huge recency bias in anything except national archival and university research libraries. And donations are not put on the shelves, they go into the booksale pile and then the pulping pile. Collections are almost always only added to via librarian purchases, procuring new books at their discretion and patron request.

A few put them up on third-party bookseller sites for a time, and I've been able to get a lot of rare and notable works for comparatively little money. Actually, a lot of them come from university libraries, so maybe they're not as much of an exception as I think.

> Why keep books, anyways? You can just ask the AI to re-generate whatever it is you want to read about on the fly

Surely, there are multiple sci-fi stories exploring this exact premise. I would love some recommendations in my replies.

The argument seems to be based on the premise that books are picked up at random from the pool of all books in the world. (As if what constitutes a book is even well defined to begin with.) But his encounter with The Book of the Dead wasn't random. It's a classic. Such a renowned book in fact, filtered by time and intellects, that it could be found even in the library of a small Christian town in Ohio.

It's a time investment for everyone involved to sift through enough cruft for these classics to surface. Generating orders of magnitude more of increasingly plausible cruft means that the human arbiters can spend a much greater time investment attempting to surface nuggets of human wisdom that simply aren't present. How do classics appear out of that environment?

The signal-to-noise ratio in books of any kind was already quite bad. I used to read many books every year but now it's less than 10. And one third I don't like anyway. I got tired long ago of crawling GoodReads and all the forums. I don't know what is driving a race to the bottom but AI had nothing to do with it. Perhaps the attention span and literacy?

In fact, I hope for an LLM service where I can schedule books as good as 10+ years ago.

Why would I buy some low-quality generic book on Amazon/Audible/etc? AI generated or not.

I was recently at a lecture where Ted Chiang talked at length about this. He said there are already so many books published every year that he doesn’t think it will matter. The talk was moderated by a bookseller who agreed vehemently.

https://www.geekwire.com/2023/ai-chiang-bender-wishful-think...

I think the problem is curation more than generation. Amazon could do a much better job in general with poor goods showing up in their marketplaces.

Some sobering statistics here

https://ideas.bkconnection.com/10-awful-truths-about-publish...

Summary: over 2 million self-published books each year in the US alone.

Now think about blogs. But as a reader, I don’t find this much of a problem. Yes, I rely on book reviews and recommendations by friends and on author name recognition. But that seems a lot better than relying on advertisements and Amazon’s algorithm anyway. And it’s pretty much what people have done for centuries with bookshops and libraries.

I think you can kind of view this as a Signal:Noise ratio problem. Previously there were good books and bad books, but the signal to noise ratio was relatively high, it wasn't particularly easy to write passable bad books and the bad books were a lot worse than the good books. So it was very easy to distinguish the good books from the bad and there were enough good books. Now LLMs have made bad books better - making it difficult to distinguish bad books from good books essentially raising the nosie floor, and made writing bad books easier - increasing the total noise in the system. This results in a much worse environment - it's much more difficult to pick out good books from amongst the noise and there's much more noise. Imagine someone rolled out an extremely popular new product that let people broadcast across the entire nation on FM radio frequencies with no way of stopping it. Suddenly FM radio is flooded with shit. What do you think would happen? People would just abandon FM radio entirely. Radio stations would give up - they can't get to their listeners anymore because they're constantly getting over-ridden by random people broadcasting, and listeners would turn off unable to listen to anything.

I think the same thing may happen here - some genres may literally just disapper because there's no way to match readers with writers anymore in a way that's economical.

I think there is still a place for filtering of physical books. Both bookstores and libraries do a reasonable job of presenting a set of potential books to read.

I'm reasonably happy with the average physical book I read. That's much less true for ebooks.

And I mostly buy books from Amazon that I've heard about elsewhere so don't depend on that filter as much.

Physical bookstores are on the complete other end of the Signal:Noise ratio spectrum, you've got such an enormous cost of publishing that even quality isn't enough (and this is the old problem of book publishing) you won't get a physical book produced unless (a) you have a track record, (b) you publish it yourself or )c) you're a celebrity who is using their celebrity to sell books or (d) you're great and you sign a usurious deal with a book publisher.

In the same way that the noise floor is becoming too high for eBooks, the quality threshold is too high for real books - (d) basically doesn't happen because you have to be so great and sign such a bad deal. Not least because publishers have consolidated into an effective monopoly.

It will be great for editors and content curators - separating the good from bad, genuine from synthetic. This will further emerge and be valued.

Reputations as “reliable quality content go-to points” will be made.

> Reputations as “reliable quality content go-to points” will be made.

Can you expand on this? Are you saying people will look for publishers like o'reilly in order to find reliable quality books? How is this different than today’s landscape?

Exactly:

Today you expect the likelihood of a quality book title to be much higher from O’Reilly than, say, Packt Publishing.

With the advent of AI Generated Books, reputable sources will count more and more than they already do.

sure, nice dream. Dream on! but there are some practical aspects that you totally ignore in this PollyAnna summary. Categorical listing of "problems" can go on for pages, and will be incomplete. Perhaps a few to start it off?

* discernment requires effort.. someone has to make an effort, and do it effectively. In the city, time is money. Few to none in the "competitive retail markets" will seize the day to waste, making no money in discernment. Hollywood knows this for decades, and has many many ways to deal.

"She knew a good movie from a bad one, and knew how to sell both" was an epitaph for a crooked Hollywood agent, who was murdered by mobsters in her 50s not long ago...

* Publication incentives.. as noted many times here today, those who pay to advertise, do not have the incentives of discernment you are so confident in.

* Security by Obscurity -- in a blizzard of new publications with new incentive and cost structure, you will apparently find the perfect snowflakes to promote to other snowflake lovers? maybe.. for a minute.. but look! more snow

source: years in the book industry in a big University town before the Internet

I agree and hope for that as well. There is a chance that someone will start a "quality market place" with a high barrier to entry that may not be particularly comprehensive but "good".

Sure there would be the usual shenanigans when it comes to gatekeeping and objectivity but in the end it could still be valuable.

To be fair, this was the status quo even before AI. Arguably word processors were already a serious source of low effort garbage.

It's weird, I always expected AI categorization and search to arrive long before generative content. But despite a lot of research going into that direction, it never really arrived in the consumer space. Search today feels not fundamentally different than search 20 years ago and if I want to know about a movie I still use IMDb like it's the 90s.

The already enormous mountain of content out there keeps growing, fuel by AI now, yet ways to explore that mountain keep staying the same or even diminishing, as the clutter in the results just keeps increasing.

I think the issue is LLMs don't scale well to the level needed for search. They also have obvious latency. So generative usage seems to be the best fit at the moment.

But you could feed the text through the LLM, generate a summary and a list of keywords, characters and locations and search through that the regular way. And as far as I can tell, that doesn't really exist yet.

IMDb has a keyword feature[1], but that's still completely human-curated and quite useless due to that, as it just doesn't have the necessary depths and completeness to discover interesting new content.

[1] https://www.imdb.com/search/keyword/

It may kill the market of books, if it weren't already badly damaged by publishers and editors. A lot of pre-AI books, specially on the latest years, were of bad quality, even recognized authors wrote bad books, but very extended, seeming more worried about selling words and pages instead of quality content. There were a lot of garbage already, and somewhere floating in that sea, some good books.

Curated books, even if they were written by AIs, could be a way to get out of this cycle, at least if editors/publishers didn't had all the incentives to lay a hand on it (Goodreads is a good precedent for this).

Another way of get out of this may be to turn over the economics of books. Did you read something good, that enjoyed and considered that it was time well spend? Then consider paying for it. With digital distribution the cost of having more readers is nearly zero. AI or human written books, what in the end matters is how it was the experience for you. And maybe how much you trust in whatever made you to pick that book.

At the start I thought that the article was about killing books as in the experience of reading books. Having AIs that somewhat had read already the book let you have a summary of what is discussed there, even have a discussion and analysis on the content, maybe even posing as the author or the main character or an expert on those topics. You may not "need" to read the book itself, and decide for a shorter activity. That won't be the end of books or reading them fully, but for some books, some topics, one approach may be better than the other. That may affect how books are written, or what are exactly books from now on. And shorter fiction like articles, blog posts and so on.

I was trying to get back into reading, and after re-reading a few books I remembered liking I went on Amazon and sorted Sci-fi books by popularity or rating or something. Basically nothing on the first page was classics, it was all novels I'd never heard of with AI generated covers.

I dug a little into one of them and it sounds like it was an independent author (they posted about it on reddit) who I guess didn't want to put the same effort into procuring a cover as writing the book. A lot of authors I think don't have a lot of respect for visual arts and kind of see the cover as a forced labor to publish a book. TBH sci-fi book covers with abstract spaceships and rainbow nebulas are one of the easier things for AI to believably churn out.

But I guess I kind of use the effort put into the cover as a way to gauge how much the author and publisher themselves think the work is worth. Even if I could be sure that the books weren't AI generated themselves (I can't) I left thinking, yeah, I'm probably never going to read again, because I have absolutely no metric, however bad any more, for guessing about the quality of a the book.

I do this for blog posts, articles, emails, etc. If i suspect in any way they are AI created I close it immediately.

The only reader AI deserves is another AI for the marketing side.

> I'm probably never going to read again, because I have absolutely no metric, however bad any more, for guessing about the quality of a the book.

Really? This might be the laziest thing I've read on HN. 30 seconds of Googling (DuckDuckGoing?)

https://www.thehugoawards.org/hugo-history/

https://nebulas.sfwa.org

https://www.sfadb.com

https://www.esquire.com/entertainment/books/g39358054/best-s... (and tons of other lists like this)

https://www.goodreads.com/list/show/19341.Best_Science_Ficti... (etc...)

Yes, searching on Amazon might not be the best discovery method anymore, but there are also institutions that curate.

Ah yes, the Hugo award, known for such quality and revered works like “Space Raptor Butt Invasion” by Chuck Tingle.

How can I trust any award as a standard of quality that has ridiculous stuff like that getting nominated?

Nominations aren't awards...?

I've read a number of books from those lists in the past and they're not really what I'm looking for, and a lot of the books I really enjoyed are not in the circles those lists select from.

But I guess more to the point in the past I've grabbed books with interesting covers or titles off the shelves at a library and enjoyed them, and I don't feel confident in that method working today. Or at least, I feel like there was a time when the rankings on Amazon were probably a lot closer to what you'd see in libraries.

Kind of a weird reason to not read anymore. 1) I never even thought about sorting by Amazon ratings and 2) I never paid to the cover myself.

While I am sure it shares metrics, I have usually had success looking at past books I have read in goodreads and their related recommendations.

as a tangent, I think the issue with Amazon goes way beyond just books. At this point it's just the American version of Wish imo.

This is definitely a case where you are much better served by information curated by enthusiasts and communities. For science fiction and fantasy, there are a bunch of images out there with flow charts and recommended entry points which should help you figure out what you like¹. And of course other actual people often have their own recommendations, as do independent book stores.

1: Here's a blog post reposting an older one with a lot of the common classics: https://blog.zog.org/2017/06/nprs-guide-through-the-top-100-...

> I kind of use the effort put into the cover as a way to gauge how much the author and publisher themselves think the work is worth

If only there were some sort of short and catchy idiom that encapsulates some wisdom about the potential folly of judging the interior contents of a book just from looking at the exterior coverings.

AI generated book covers are also a thing at this point, which reinforces that age old idiom. Even legitimate writers will still find value in generating their covers.

"The Amazon algorithm didn't show me the best book in the world as the first search result so I'm never going to read again".

Something tells me you just aren't interested in reading (which is fine) and are trying to find excuses for it.

I read lots of sci-fi and this story doesn't sound right.

Are you sure they weren't just modern sci-fi authors you haven't heard of?

Sci-fi has never been stronger and it makes sense the most popular wouldn't be what you read way back when.

That's doesn't mean they are written by bots.

The main thing AI does is kill off blind purchases of ebooks from unknown authors. Reputable publishers won't be publishing and bookstores won't be stocking AI generated books so it's mainly an ebook problem.

The way to successfully write a book as an unknown author without a major publisher intent on making you famous is already to build a social media following[0] then leverage your social media following to promote your book.If nobody's heard of you, self publishing a book has always been a waste of your time. It's just even moreso now because somebody is using the BS generator to write fake books and flood Amazon with them.

[0]: This doesn't just mean writing social media posts but could also involve things like getting published by magazines, doing podcasts, appearing on TV or even just doing Twitter Spaces. The point is to be a known figure by your target audience before you write a book.

Yes, in the next few years the entire pop book industry is going to decimated. Just like the nearly gone magazine industry. However, what evolves to replace it is going to be very interesting. Humans are social animals, we're not going to stop communicating. Confusing prose? Will become a relic. Books that explain how to use some product or tool or adopt some behavior will become relics. LLMs will provide that learning resource. A hybrid LLM that tells a specific, intellectually challenging, Nobel quality narrative is what I hope is the result; it would be like talking to the survivor of an epic adventure, and you can both listen to the narrative as well as interrupt and ask questions. Likewise, new scientific theories would be published as an LLM that can explain at every level from Phd to 5 year old.

Tangential, but those who love the language used in the Egyptian Book of the Dead, surely know about the opera Akhnaten by Philip Glass which used some of the text, both translated and in Egyptian. (Warning, opera, and especially Philip Glass' operas, are not everyone's cup of tea. You may like it or hate it. The point here is to showcase Book of the Dead text).

Hear the opening verses in English here:

https://www.youtube.com/watch?v=hRyauSlnP54

Hear the opera scene in Egyptian with subtitles here:

https://www.youtube.com/watch?v=XENvMGyy4J8

Here's a concept: on-demand, neverending AI books that ask you for feedback (and/or take into account variables like reading speed, duration, etc.) and actively try to predict what you'll find entertaining as you read.

I can't speak for Amazon's, or any publishers policies, but there is at least one publisher screening submissions to hopefully filter out AI generated stories. That would be Clarkeworld, a science fiction magazine, who was famously overwhelmed with low quality AI-generated submissions, as was discussed on HN last year.

https://news.ycombinator.com/item?id=35999896

> authors tried to eek out an existence

* eke out

https://en.wiktionary.org/wiki/eke_out

I am hoping this gives publishers an important role of becoming the quality gatekeeper. The publisher has the relationship with the writer and ensures a certain level of quality. If a writer keeps coming up with nonsense (AI or not), they’ll cut ties. For me as someone who reads about 3-4 books a year, I don’t see an issue with this. It sucks for this indie market, but even before AI, someone had to plough through the shit.

This seems like a space with a problem to solve: identifying AI-generated fiction and rating how much of a given work fits that classification.

The problem is that it's pretty easy to fine tune an adversarial model to fool classifiers. In some sense, building that classifier is a very productive way to improve the model, so it's something the generative side is already doing. That makes it hard to win the arms race from the classification side.

But also the solution. With enough classify -> beat the classifier iteration, the works will be indistinguishable from human production, at which point there is no longer a need for classification.

This assumes that the limit is full generative parity, and not just a our SOTA classification techniques maxing out.

This post should be retitled to indicate that the author is referring primarily to "Amazon self published" ebook sector.

Authors today can't be invisible. They have to market their work, which they do by going on podcasts, giving media interviews, some kind of social media presence as well. This filters out most of the AI flotsam.

Did anyone else pick up on a condescending undertone off of this article towards people who don't like reading? Reading this felt like the author was saying "I love reading, and if you don't, you're wrong."

Regardless, this is the same argument as the one being made about AI taking artists' jobs.

As a distinct difference to personal computers replacing typewriters and the internet enabling self publishing, both of which transformed writing, AI is much faster than humans in reading and analyzing books and thus can also be part of a solution to problems from this new transformation in writing.

In the past quality was maintained because it cost money to publish a book.

The solution is to charge a nominal fee to self-publishing a book. Maybe $100. The author now has to be confident in the quality of the book and is betting it’ll generate at least $100 in profits.

And that’s why the apple store has better apps than google play. Paying $100 every year is enough to deter drive-by scams.

Misses the angle that there are people with something important to say who have trouble writing - LLMs will help get them out there. An LLM can take some poorly expressed insights and developmentally edit the text.

Self publishing is already 90% trash that you have to wade through to find a gem. LLMs will maybe make it 95% trash. So overall bad for the industry, yes, but not as game changing that people are making it seem.

If it was worth a human spending weeks writing the 90% trash. LLM make it economical for 99.999999% to be trash by next year.

I doubt it was monetarily worth it before expect for a handful of people with a following from somewhere like a blog

Most writers write because they want to write and think they have something to say. Very, very few make money. Ted Chiang made this point in the lecture I linked to elsewhere in this thread

Looking forward to next LLM model trained on books written by LLM models....

Please, give me more "AI is killing <whatever>" it's not tiresome at all.

Books were being "killed" far prior to AI.

It may kill e-books, but nobody is going to go to the time and effort to print physical books written by AI.

If you want to read books, buy real books (or join a library).

Unless the LLMs have suddenly gained API access at Amazon, the correct title for this piece is "SCAMMERS are CONTINUING to kill books".

> The book market relies on a vast army of unpaid volunteers to effectively sacrifice their time and wade through a sea of trash to find the gems.

In other words, the market already sees no value in books. If it did, the money would be pouring in for this type of work. AI isn't the problem.

AI might be the solution, though. It is conceivable that AI could find a way to make books appealing to the market in a way that humans have failed to discover.

I guess the average ebook on Amazon written by a human is already pretty close to something a LLM would generate so...

I don't like the equivalency of "books" with the sort of low-effort drivel that LLMs will replace. In reality, an LLM isn't going to replace an author that is able to communicate a particularly unique viewpoint or experience.

I do think, however, that book sales are going to continue becoming more dependent on author-as-person style marketing. Some of the most lucrative books in recent years were functionally add-on products to whatever the author's main "business" is, and the author themselves went on dozens of podcasts, etc. to tell their story. The days of being an unknown mass-market writer that mails a manuscript to a publisher then disappears, is probably over.

This is somewhere in "think about the horses" territory.

- AI will raise the floor in a lot of industries, more quickly than it will raise the ceiling. I don't see a problem with that. Less garbage is nice.

- Good writers will make use of AI to be better, much as they (and we in general) did with the internet, and so many other technologies before that.

- If we can not tell good or bad apart, then we should be extremly suspicious in how far it actually matters, specially with everything that is not grounded in physics. Truth is fickle. Feeling ambivalent about things that can not be argued away by way of physics is probably a good thing on average.

- Bullshit detection has always been an issue and will continue to be so. As far as I can tell, we have been getting better, not worse, at this (if you consider the absolutely monumental increase in total bullshit generation that we had to cope with over the past years.)

- In the end, as per usual, despite all claims to the contrary, people will not care how it was made. All that will matter is if it does something for somebody. Note that how it was created might do something for somebody, or, more likely, a certain illusion of how it was made will be good enough and more economical. Over time the sentimental power will fade.

There's already more good books than what you could read in your lifetime. Which means you have to pick them wisely. Which means good services that pick the best books for you are valuable. And: the key problem with AI here is the risk of drowning quality content in the sea of nonsense or, worse, disinformation. But it's not limited to just books but content in general. So: let's sharpen our focus to choose wisely what we consume and support content curators who preselect the best stuff for us

This suggests, more than ever, that human curation and review is absolutely crucial. The sad thing about this is that it just lends more power to monopolistic publishers and renowned critics and away from people who, without access to such esteem and networks, have to work tirelessly to produce self-published works. All in the hopes of earning a pittance and finding a way to thrive artistically within capitalism. It's strange that AI, contrary to what true believers say, may not end up democratizing information, but instead it will make us more reliant on centralised and assured sources. Perhaps the journalism industry, and other ostensibly old fashioned institutional vestiges, are about to see a renaissance.

EDIT: Tangent: I run a book recommendation platform and am envisaging having to implement a pre-2021 lock-in/time-freeze as my data is about to get massively polluted by the AI Boom.

Books - It's doesn't mean it's the source with evidence. Books also has a message has bunch of lies and illogical theories to hallucinate the readers. Sapiens is one of the greatest fools book.

AI is really helpful in this case, reduce technical books, gives very short and sweet answers -> This all we need either book or AI.

AI can't become Adolf Hitler, so the view point of the Author still remains. AI can't replace it.

The article is mostly nonsense mixed with misunderstanding. There was a time before publishers where everyone essentially self published by paying a printer, that never really went away but what changed was advertising and distribution. If all a publisher did was print your book then it'd have the same chances of being read as if you had self-published. eBooks have had a mixed impact because they a) lowered the price and difficulty to self-publish and b) made wide distribution cheap or free but they did not have any meaningful impact on advertising. Introducing LLMs or any other form of automated production of content to the mix does not impact advertising either. All it does is increase the speed of production for everyone, good authors, amateurs and those looking to make a quick buck.

The author complains that they are finding bad content when they search for books. Guess what, that's a search problem not a problem with the content. Search is only useful if it returns relevant results. All this talk of betrayal and trust is just a symptom of crappy search or recommendation algorithms. It doesn't win any sympathy that they are also complaining about free books on Kindle Unlimited so they aren't even cheated out of money, they simply lost a tiny bit of time since they indicate they can quickly identify machine generated content.

This

> There is no feeling of betrayal like thinking you are about to read something that another person slaved over, only to discover you've been tricked.

and this

> Part of the reason people invest so many hours into reading is because we know the author invested far more in writing.

are incomprehensible to me as they appear to be some subset of sadism that derives pleasure from someone enduring a form of hardship. Not quite the same since parts of creative work are enjoyable but still weird because any form of creative work will inevitably have large sections of difficult, tedious or just unpleasant effort that goes into it. Saying that something has less value to you because its creator used a tool to make the bad parts easier to do is just wrongheaded. The only argument that could stand is if the tool they use made their output worse in which case it is justifiable to criticize it but the same goes for an author who doesn't bother to edit his own work or to ask another person to check it and edit or takes other shortcuts like ignoring consistency or using tired plot devices or copying some popular style.

> Good writing kills its darlings. If you don't care enough about a section to write it, then I don't care enough to read it.

That is just not what that phrase means. "Kill your darlings" means to throw away parts that you care about, it's literally the exact opposite of what they're saying here.

Honestly just skimming the rest it seems the author does a lot of work to paint a picture but it doesn't do much to support the argument. A great deal is made of the effort it takes to read a book and decide if it's worthy of recommending or selling in a book store. This just ties back to the search problem and is ironically a place where LLMs and similar ML tools could help a great deal since they can make for excellent classification and recommendation engines. It's pointless to complain about the volume of books since this was already an untenable problem with only human authors and the sheer weight of history. The author says that a book seller may read 80 books a year, certainly an accomplishment but absolutely nothing compared to the number of books published each year[0]:

- 500,000 to 1 million from traditional publishers

- 1.7 million from self-publishing

- 130 million globally

Once again, it's a search problem. When you have 130 million new titles per year it really doesn't matter if you make it 230 million or 1 billion if your solution is to chip away at it 80 titles at a time, you need automation. Fixing search and recommendation is the only thing that will impact the awareness and advertising side of the business. If you don't fix it then yes publishers will carry more weight for their ability to vet their authors but this is nothing new and was not meaningfully impacted by digital publishing as already established. I'm afraid the thing the author is decrying is exactly the medicine they need.

[0] https://wordsrated.com/number-of-books-published-per-year-20...

AI is already getting killed, people being disgusted and annoyed by this garbage.

Central command seems to have instructed drones to use “already” as a means to gaslight people into thinking there’s use and demand for this gibberish. Dont fall for it.

It’s essentially the next logical step in this grift: https://youtu.be/biYciU1uiUw?si=_K1VAtnj-VRXCONQ

People are posting AI grift books, but is anyone buying?

AI will come for everyone , last day i read about truck driver (copilot) will be replaced by AI and today this.

How much energy does a truck driver AI use compared to a human?

Energy is an interesting comparison. Perhaps we should look at $/energy unit, and compare those dollars to the full cost of hiring a human including time off and benefits and payroll taxes and everything. We should also consider that the human consumes a lot of energy that the AI does not: many truckers have PC's, TVs, consoles, etc in their truck and burn fuel to play video games during their off hours.

I don't think we're going to replace truck drivers any time soon, but I'd be surprised if in say 30 years the vast majority of trucking wasn't automated.

But they would do that at home as well. So irrelevant. If anything they consume less energy than people in homes.

> Perhaps we should look at $/energy unit

But then also consider the change in this price as energy consumption changes, as a result of jobs becoming AI.

Human energy consumption is highly variable. Especially when you do cradle-to-grave cost analysis. Are we talking about an average american truck driver? Or the ideal high-efficiency truck driver that is raised to only drive trucks, grows on agaragar and undergoes early depowering after its working years?

How much energy does a human use in 18 years of training?

There was a very good talk at NeurIPS this year that was comparing LLM training to amount of energy expended in building wonders of the world. I think this would’ve been a good thing to include too.

A LLM trained once can be instantiated a million times, while each human instantiation requires individual full training.

no it isn't

(NM I'm an idiot)

Did you read a different article? That is not what the author wrote about.

Reply on news.ycombinator.com