humans. humans did this. not an ai. not some cold machine. cold humans.
the simplification/corruption of discussion through misuse of “ai” has been happening for decades. the shift is that now it is being used to shield humans from responsibility. that cannot be permitted.
This is what baffles me about the discussions on AI making the search based experience of the web worse. Worse? Have you searched for anything in the last 10 years? It’s just awful SEO garbage - at least the LLM SEO garbage might actually tell you how to pickle dog biscuits, while the human generated garbage just meticulously structures the text to imply they might tell you while they actually can’t because the don’t actually know.
But regardless of whether it’s LLM SEO or human SEO, a human decides to make fake pages about pickling dog biscuits because it was more important for them to make some money personally than for the search based internet to be anything but garbage. A human did that. The same humans that have been screwing everything up for everyone else for at least the last 10,000 years.
IMO I think this trend is just an acceleration and democratization of enshitification tools. But I think the future result is three fold:
1) dependence on automated algorithms to differentiate between garbage and quality is now demonstrably dead, while it just seemed dead before
2) open LLMs backed by information retrieval systems will be a much better tool for learning than a naked search engine
3) humans will need to curate lists of good material and depending and algorithms and automated information retrieval systems alone will be insufficient, and hence curated and editorialized directories will become the primary way of discovering human generated content of quality.
> a human decides to make fake pages about pickling dog biscuits because it was more important for them to make some money personally than for the search based internet to be anything but garbage
If I think I make the best dog biscuits and I want to be found, then it’s not a “fake page” from my perspective — it’s just marketing and in a way I’m doing you a service because you’ll be happiest if you buy my dog biscuits.
Yea, this is the story that marketers tell themselves so they can sleep at night: "Well, I'm different! Unlike everyone else, my product is high quality! People will want to see it!" All ~10M marketers in the world excuse their actions with this rubbish.
Not necessarily, but even if so, the fact that all N million minus one of them incorrectly think they are the special one, results in the sea of crap we have today.
Another tough pill for marketers to please to swallow: Even IF the product is great, AND it's exactly what I need, AND I would be happy to buy it... I don't want to hear about it randomly throughout the day as I'm trying to live my life and do normal things other than "discovering products". I don't want to be actively or passively contacted uninvited by companies, even if I would be interested in their products. I'm not shopping. I don't want to hear about it.
That’s the power of a curated list. Consumer reports, Michelin, etc. It’s not you that’s advertising your product, it’s a trusted third party. Gaming search engine scoring algorithms is not the right way to prove your product is better. It’s the way to game a search engine to force yourself on to people.
Do you think you’ll be found? That’s not how search engines work. You’ll make your lovingly made page, but someone with a link farm will bury you 20 result pages deep. The internet as indexed by search engines hasn’t worked for at least 10 years.
And no one will find your page with the best dog biscuits because it’s buried on page 30 of the search results behind all of the people trying to generate ad money who don’t actually know anything about dog biscuits
That’s a different scenario altogether. I think the parent is referring to link farming and arbitrage type pages that serve the purpose of netting a commission, not selling more first party products.
this tracks, and makes an interesting argument for a modernized librarian curriculum. maybe there already is something of the sort under a different name. llm usage makes a lot of sense for dealing with… data.
granted, universities are decaying in much the same way as information access online.
Right. AI is a tool utilized by humans. Humans did this.
I'm increasingly concerned about how regularly I'm seeing examples of entire framings of issues that serve to externalize responsibility. AI is a very fascinating, arguably unprecedented kind of tool. But it's just a tool, wielded by humans, to do the desired bidding of the humans, who are in complete control of that tool.
A different POV might be: humans built a tool that, because of various extremely difficult to change propensities in human cognition, it was obvious that humans would find it hard to control their use of.
While this had been true for many previous tools (especially during the computer age), the fact that this tool evoked one particular human behavior - taking the intentional stance [0] - created many additional problems.
I'm sure this approach—just regulate the symptoms without thinking too hard about the underlying causes—will yield just as effective a solution as it has with guns. We're doing a great job, let's keep up the momentum.
before regulation, I think we have to settle on the moral aspect. It’s not clear for everyone for guns, so don’t rush for «AI» if you want effective results, regardless of the projected risks and the urge that goes with it.
Few days ago Democracy Now! on YT had a first-person interview about corporations, taking advantage of difficult situation, trying to buy out land from people who owned the land for generations. Can't find now in which video it was.
the human condition has not been addressed successfully. both the creation and mis-utilization of explosives (or similarly explosive “disruptive technology,” as misused as that term is as well) seem inevitable.
bombs are just controlled explosions used for evil. but sometimes it benefits humanity to move mountains, and explosives make that easier.
a daunting task, but not insurmountable. oddly enough, language models could be utilized towards making this an easier problem to solve.
but you aren’t wrong, and the “books” on amazon are proof. misuse already rampant.
I think you’ll find that people are more willing to take your posts seriously when you use proper capitalization and punctuation. Particularly in a place like HN.
The author is Dr. Miles Stones, Miles Stone! Milestone!!
As in, whoever made this (using Ai) considers it a milestone in someway.
I can't have been OpenAi out of the box as it doesn't have the data, but it'd be quite easy to load in all the news articles about the situation to a custom model and then have it generate the book.
Yeah, I'm giving them too much credit and I was presuming you'd want to collect every piece of content written (news articles mainly) and that would be the context, but it'd be too much data for a prompt and probably not needed to create a book. I was thinking the book would actually be full of real world data and real commentary etc.
You could probably just ask GPT to come up with decent chapter headings for a book about a recent large wildfire and how it impacts climate change policy.
Then for each chapter heading give it a couple of pages of input info and ask it to write a chapter based on the content and the heading.
Bet it's a really bad book if this is even close to how it was made.
As an AI language model trained on data from before September 2021, I haven’t personally read this book, but based on its features and customer reviews, I can confidently give it a five-star rating.
Sounds like a "race to the bottom" on quality of information and content. Has crypto bro considered putting their efforts into a real job or a real company instead of shortcuts that ultimately are a net negative to society? They'll likely be more successful and probably even feel better about themselves.
Would "bestseller" be influenced by people grabbing this BookSpam through Kindle Unlimited? It looks like the bulk of the Kindle best sellers[1] are available through unlimited, and I have to think Amazon counts that as equivalent to a sale, even though it would largely be people who grab low grade junk like this that they would never have paid for traditionally.
So are people dumping this junk to yield royalty payments through the Kindle Unlimited program? How does that work? Minutes read? Do you get paid just because some person downloaded something via the program, and if so how much? Time to fire up the text generator.
That's assuming it is actual people reading it, and not just click fraud. Amazon has been battling that for years, books filled with literal gibberish reach the top few hundred titles in a category and the only reasonable explanation is that some group(s) are gaming the system by buying/phishing Kindle Unlimited subscriptions then skimming through their own "books" to turn a profit.
Using LLMs to make the books vaguely coherent instead of meaningless word salad adds a layer of plausible deniability which might make Amazon slower to notice what's going on.
It seems like a kind of strange strategy for click fraud. A KU subscription is $10, Amazon takes a 30% cut, and for $700 a month you would need 100 KU bots with unique accounts and payment info.
I don't think this "book" filled the pages with loren ipsum, but it is entirely possible that they fed all of the available news stories into a language model such that someone actually sat there page flipping through chapters of decorated text.
But ultimately I think there is a stark difference between someone who actually paid $7 for an ebook, and someone who downloaded something on Kindle Unlimited and flipped through it. I am curious how KU influences the best seller charts numerically, which I can't find a clarification of.
How hard would it be to write a script that downloads books and "reads" them -- i.e. exercises the Kindle app (say on a desktop-alike) -- giving Amazon plausible excuse to say "n words read"?
Point being: if there's a way to game a system, people will game that system.
> I suspect that anything that rises to the best-seller level gets examined by Amazon very closely.
I think you’re vastly overestimating how much Amazon cares.
See their physical goods store: it’s also a race to the bottom with zero oversight whatsoever (commingling, counterfeits, used/returned/dirty items sold as new…) I don’t see why the virtual books store would be any different.
Amazon wrote the proverbial book on enshitification/race to the bottom :(
"Seller" definitely seems to be a misnomer here. That said, I'm not so sure it's out of line with publisher tradition as units shipped is often used to hype (alleged) demand. The music industry had the same problem (along with payola).
The more things stay the same, the more we really don't change.
This might be better with a title that indicates the red flags are about Amazon's quality assurance. Initially I thought it was a post along the lines of a "How could a book be ready so fast unless the wildfires were planned??" kind of conspiracy theory.
You either have gatekeepers--publishers, agents--who bring some level of quality control or you throw all that out the window in service of democratizing access. Not that they would, but Amazon could absolutely only accept books from legit publishers but can you imagine the howls of protest here were they to do so?
I'm not at all sure you're wrong. Or at least there may increasingly need to be some reputational authority whether it's as a person's real-life identity or authoritative third parties. You sure can't trust reviews or self-written puffery.
If one is cynical enough, it's easy to imagine shifting back to a world where you have a relatively small number of (more-or-less) trusted media outlets you have to subscribe to, some big publishers, and everything else is sort of hit-and-miss and most people won't bother with.
Whether real or perceived, this highlights why I buy almost everything I need in a physical store; there is an implicit statement that the product is good enough for the store owner to carry. (I also avoid chain stores, I basically only shop at small, locally owned businesses... and Aldi :D )
I see your point but need to say that I find the curation of B & N to be decent, including the "technology" section -- you can find a good mix of beginners and advanced level books, and they are often up-to-date.
And slightly off topic -- libraries (both public libraries and university libraries) can sometimes be exactly opposite of this. Whenever I see people say "find a book from your local public library" I know they never been to one and carefully check out the catalog themselves. New books don't arrive on the shelves until several months or a year later, and libraries rarely get rid of outdated books that teach you to learn Windows 7 or C# 6.0.
That is a very astute observation. Bookstores curate based on what they expect to sell and they take on risk when they order books ahead of those sales. But that also is a bit of a self-fulfilling prophesy: you can not buy books they don't have for sale. And more than once I've bought a book that had been re-priced multiple times. We even had bookstores in NL that would carry leftover stock from other bookstores (notably: De Slegte, where I spent many hours sifting through the stacks looking for gems).
> they take on risk when they order books ahead of those sales
I think this varies by country, but in the United States most books are "returnable", meaning that the publisher gives you full credit if you return them. In addition, publishers often don't require payment of the invoice for quite some time after the books have been received. There's little or no risk involved, other than the opportunity cost of not having the shelf space for something else.
That's for hardcovers. For paperbacks and periodicals, often the publisher only requires tearing off the front cover and returning that. The bookseller is supposed to destroy the rest of the publication, but unscrupulous ones often launder these so-called "stripped books" through third-party brokers.
So, consignment sales. That makes sense I guess and would obviate a part of the GPs reasoning, though there still is risk involved in that the stores would have to figure out how they allocate their other scarce resource: space.
Inspired by this discussion, I decided to see if I could get Anthropic’s Claude 2 to write a similar book about an ongoing news event—the tropical storm that is (as I write) approaching Southern California. It did very well—good enough to be published, I think, as long as nobody checks the facts. Here is the result:
> It did very well—good enough to be published, I think, as long as nobody checks the facts.
After subjecting myself to actually reading that, I wouldn't expect it to do well once somebody also checks the prose. On every iteration and in every variation of generative textual production tech I've seen so far, I'm increasingly encouraged with the thought that the only writers who will be negatively impacted are the ones who really should be.
Who is that human? Is it a singular person? Is that person paid by the store that has incentives to promote an inferior product with higher margins than a better product for less margin? So many questions which leads me to one of my own…what do I find as a reputable reviewer? At this point in the game, I have so little faith in reviewers, I don’t know how to answer that honestly. Even a site dedicated to “honest” reviews would immediately be suspect in my mind. How are they making money? Who is paying for the items they are reviewing? These types of questions are the ones that always give me pause in accepting credentials
exactly. I worked at FB and commuted with some folks on the CP/dismal/abuse review team (they reviewed contnet flagged for all sorts of crazy shit)
It was disturbing.... I wouldnt want to be that human, nor know those humans as well - there is a lot of dark content that makes ignorance bliss - both from knowing the content and knowing the humans (even the mods) around it...
I write very fast but that is seriously impressive. Wow. My record is a 25 page single spaced report in two marathon days. I also took the week off after that. The customer was in a hurry and we made it work but normally it goes quite a bit slower than that. When writing fiction I guess it is a bit easier, you are not going to be raked over the coals for some subtle contradiction. But still that is a massive output.
Have to say that when writing fiction (and taking that seriously) my output slows down by a very significant factor. It's not the physical act of writing that takes time, but the time needed to think about the scene, characters, etc. I've said before that I never realised that writing (fiction) is a performance art. The performance might all be solely in your own head, but it's as real as any, and harder, because it has to make sense on the written page, in light of all that's been written before it and has yet to be written after.
eta: Writing fiction I consider myself doing really well if I achieve 500 words in an hour. Non-fiction? 1000 is quite easily done.
That's interesting. And it makes intuitive sense: when the facts are what they are there isn't much that you can do other than to translate those facts into good prose and to try to accentuate that which is important and to de-emphasize that which is not. But when you are constructing a whole world from first principles or when you have to imagine a series of events that have to be plausible it gets harder. My job is mostly constrained to avoid writing fiction!
I have never tried my hand at writing fiction, so I probably should shut up about that :) At the same time, I told / tell my kids bedtime stories that I pull out of my fantasy at the drop of a hat and they can go on for as long or as short as I want. But those are not to be compared to anything novel (or even novella) sized. 15 minutes tops.
For non-fiction, I won't say I routinely do it but if I know the topic area (no real research needed), have a bunch of quotes, have some introductory boilerplate to introduce some topic, some basic facts on e.g. a product announcement, I can crank out 1,000 words of serviceable prose in certainly a couple hours. Probably can't scale to more than about 3,000 words per day--and wouldn't want to do that on a day-in day-out basis--but I have friends who are faster so 10,000 words per day is not totally out of the ballpark.
When I was an analyst, my benchmark was about 2,000 "memo grade" words per day--typically mixed in with some phone calls, etc. More carefully constructed reports etc. took longer especially if research were involved.
I just did a check on that document, here are the stats, apparently it was a few more pages than I remembered:
Pages 31
Words 17212
Characters 97347
Characters excluding spaces 80354
Unfortunately I can not find exact statistics on the writing activity. So that's roughly within the ballpark of your maximum. I was pretty happy with this particular report and so was the customer, but to do that kind of thing sustainably would be beyond me.
Edit: and, more so because this was before COVID and that hit me pretty hard.
That's a lot. I could probably do it given the right materials in a couple days but it would really be pushing things and, like you, I'd be taking a break afterwards.
I seem to recall doing some math and finding out the limit for some models of LLM was pretty close to that. If so, this book is likely the result of a single prompt, which seems to me like a lazy way to be lazy.
Extreme outliers arent useful to the discussion. If all people were capable of outputting this volume of shit at scale, then LLMs wouldn't be changing the game. But people aren't, and LLMs are a net enshittening force.
> Extreme outliers arent useful to the discussion.
Your opinion.
Gibson was an outlier, but not an "extreme" outlier. Isaac Asimov wrote over 500 books in his lifetime, all of which were of reasonable quality. Enid Blyton wrote over 800 books. Barbara Cartland wrote 700. R.L. Stine is well over 400 and still writing. There are many others.
I know the SEO battle on the Amazon search results is very cut throat, with existing entities taking a loss for weeks/months to establish their 1-3 ranking. I never expected somewhat similar SEO optimization for books, I always figured there were more checks.
I mean a more deatailed comparison, not just he full text of the book but gather up however many hundreds of articles have been published too; and then do a textual analysis to see just how much variance there can be.
if my use of the word "amusing" offends; i apologize; I am not a sensitive person in that way and frequently do that. Nothing about this tragedy is amusing.
The predicable banal venality of people's responses, on the other hand, is. If you can't laugh at people jumping to profit of tragedy like the author of this book, what can you do? Despair?
Wait, what? Isn’t this similar to saying “your honor, my neighbor does unethical things, so I do too”.
I don’t trust any big media these days, precisely because they lie, misrepresent, hide and blatantly engage in propaganda. I am not alone in not trusting the big media publications, a big chunk of population doesn’t trust either. This is not good for the society.
“If I don’t do this, someone else is gonna do it” or “someone is doing already, so what’s the big deal if I do it too?” is a very weak argument for doing bad things
the simplification/corruption of discussion through misuse of “ai” has been happening for decades. the shift is that now it is being used to shield humans from responsibility. that cannot be permitted.
humans did this.