Hacker News new | past | comments | ask | show | jobs | submit login
Mushroom pickers urged: Avoid Amazon foraging books, appear to be written by AI (theguardian.com)
144 points by sandebert on Sept 1, 2023 | hide | past | favorite | 74 comments



It seems to me that many books on Amazon are now AI generated trash. I received a book as gift for my son, purchased from Amazon, that was completely nonsensical. It was about caves. Here’s a quote:

“Some of the most common animals found in the caves are stalactites and calcite, while others are caves.”

The whole book was absurd sentences like that. Every page was a single paragraph accompanied by a random picture of a cave on the opposite page.

The “publisher” has hundreds of “books” about every topic imaginable. If you search “{topic} for kids” one of their books would come up. These texts have no redeeming quality or value. Their mere existence is a net-negative for society.

Amazon is selling this trash as if it were a real book. It’s predatory at best.


Reminds me of an obituary allegedly written by ai:

Brenda Tent retired from living at the age of old, surrounded by family and natural causes. A librarian from birth, Brenda was an avid collector of dust. She had a sweet heart and married her high school. She loved having hobbies and helping her sons to be disadvantaged youths. She had no horses but thought she did. The church gave her a choir because she sang like bird and looked like bird and Brenda was a bird. She owed us so many poems.

The funeral will be held in 1977 at heaven. In lieu of flowers, send Brenda more life.


Writing in The Mirror two years ago, OliviaRose Fox traced [1] this humorous obituary to a 2020 book by Keaton Patti, I Forced a Bot to Write This Book: A.I. Meets B.S. [2]

[1] https://www.mirror.co.uk/news/uk-news/robot-writes-engineers...

[2] https://www.simonandschuster.com/books/I-Forced-a-Bot-to-Wri...


Sounds like GPT-2 output, which is definitely the most hilarious kind of LLM output. Comedic value of GPT-3 and up is sadly much lower.


Every time I see this, I try to make it as far as possible without laughing. Today I made it to "She had no horses but thought she did".


> Some of the most common animals found in the caves are stalactites and calcite, while others are caves.

Things like this will prepare your child to the world we are moving into



And chances are Amazon is even printing it! They took their marketplace approach to publishing: allow anyone and anything to upload any crap (or other peoples copyrighted material) and sell it immediately, with no oversight or quality control.


Please name and shame.


OP did - "Amazon". That's the company that people are trusting to be offering real products, that isn't actually doing much in the way of QA, and instead just abusing customer labor to do QA and a "no fault return".


How do you do QA on one-off items or reliably fact check books?


QA involves more than just testing - verifying vendors, checking their bona fides, and having relationships/terms that allow for more than just no fault returns for fraudulent suppliers.

Amazon basically has the same broken dynamic as other centralizers like Youtube. They want to promiscuously accept suppliers as if they themselves are just a neutral conduit, while also performing discovery/recommendation across those listings. These are two very different things, and the conflation of them makes it so the barrier to getting in search results / recommended is an afterthought. If the fraudulent books were only reached because OP had come in through a link recommended elsewhere, there wouldn't be too much to take Amazon to task for. The problem is that they're discoverable through their own branded search, hence the blame of quality (or in this case, abject fraud) landing at Amazon's feet.


> How do you do QA on one-off items or reliably fact check books?

There’s a reason that traditional publishers don’t just take manuscripts and print them without active review and engagement, and there’s a reason traditional book retailers have buyers. Amazon (both as a publisher and retailer) and other online marketplaces have treated it as if the only reason for these things was resource limitations of the traditional marketplace (whether imposed by printing, etc., costs for publishers or inventory and shelf space concerns for traditional retailers) but it also goes to customer trust.

Does it scale? No, but if you don’t do something like it, then you may have an infinite number of titles, but it will become impossible for customers to find anything that has quality among the trash.


Presumably someone would read the books. Normally, this is done by the publisher, which decides which books it wants to publish. The bookstore decides which publishers it wants to work with.


https://www.amazon.com/stores/Bold-Kids/author/B09NLGSGXB

Have a look at the books that have reviews. Some include pictures or excerpts. They’re all complete trash. I found a photo I saved, here’s another quote:

“The caverns are home to many creatures, including raccoons and blind salamanders.”

So close.


I noticed most of these books have no reviews. When they do have reviews they have no text with thr review. I think we can do the most good here by just adding a one star review for each of this sellers books, with text, saying that it's AI garbage. I've done a few but I want to finish my paneer now.


Ohmygod this is brilliant. "Switzerland A Variety Of Facts Children's People And Places Book"

https://www.amazon.com/Switzerland-Variety-Childrens-People-...

Switzerland is a landlocked country with no coastline, but it borders five other countries. This peaceful country is home to one of the lowest murder rates in the world, at about five people per 100,000 residents. Its people have always been known for their benevolence, and despite the fact that it is one of the wealthiest nations in the world, the country has a reputation for being a peaceful place to live.

From the descriptions of other books:

Until the Myceneans conquered Crete, there were no toilet papers.

[Helicopters] also have limitations, such as the fact that they cannot fly in the dark.

A great way to get your kids interested in cows is to teach them some fun facts about them. For example, did you know that they can't walk up and down stairs? That's because they have no legs. Or that you can name a cow? Or that cows are only found in the United States? Here are some other cool facts about cows that will interest kids. Let's take a look at some of them.You can find lots of interesting facts about cows online. For example, cows can see in 360-degrees, which helps them stay safe from predators. They can smell something up to six miles away. You can also find out how long a cow can hold its breath. The most interesting fact about cows is that they're social animals. Even if they're not in the middle of a herd, they'll stick together no matter what.Did you know that cows can hear higher frequencies? Despite their small size, a dairy animal can make over 90% of the world's milk supply. While a few thousand years ago, milking machines were invented, and today they can milk over a hundred cows an hour.

Children are increasingly interested in climate change and the effects it has on their lives. For example, they can learn about the dangers of El-Nino

[The Ancient Egyptians] agricultural techniques also allowed them to create useful items such as pottery

Unlike other cats, however, mountain lions seem to be particularly attracted to kids. While they may not be particularly dangerous, they do seem to perceive children as easy prey. Small children can make a great threat to mountain lions, so they will usually try to run away from them. That said, running away from a mountain wolf or cat could trigger a lion's instinct to chase after them.

D-Day for kids is a great lesson for the entire family, and it's easy to see why. The invasion began at 6 a.m., and by the end of the day, all but one million troops were dead or captured. The battle was the first in the history of the Allied Forces, and it changed the perception of what a soldier could do.


Seriously. It's not libel if it's true and OP would be doing a favor.


What does a "book" really mean in the current era of online marketplaces that accept (and promote) self-publishing? A book is now just a differently formatted PDF file. It's not that different from any web site. Calling something a book conveys no authority.

In the past, getting a physical book into your local bookstore meant that someone had to spend money to physically print the book and have agreements with bookstores to physically put the book on their shelves. This threshold prevented garbage books from being made and distributed, because you had to convince someone to invest in the book and distribution.

Of course, there were always self-published books available, but they wouldn't be in the same locations as other mass-market books. You could physically inspect the book and realize that it was lower quality. These AI generated books are the equivalent of a bunch of papers stapled together.

The Amazon "bookstore" has become the same as any other "App store" - full of garbage, to prey on less discerning users and vulnerable to the same kinds of manipulation.


Yes, I recently bought a vegan cookbook just to find it full of AI nonsense, encouraging me to try it with beef or chicken. If you are looking for a new book on any topic, don't start on Amazon. Ironic that Amazon, originally a bookstore, can't be trusted to sell real books.


> Ironic that Amazon, originally a bookstore, can’t be trusted to sell real books.

They were never particularly trustworthy as a bookstore, but they undercut B&M stores on prices, so if you used a B&M bookstore for discovery, you could turn around and order at a substantial discount from Amazon.


Can they be trusted to sell real anything?


Probably not, even the fakes are being faked these days.

There's a famous art faker whose fake art is being faked. https://en.wikipedia.org/wiki/Konrad_Kujau


Regardless of the AI detection score I wouldn't trust a random book on Amazon where safety was involved. Even before LLM spam there was low quality self published garbage, algorithmic or otherwise.


> low quality self published garbage, algorithmic or otherwise.

To elaborate on this - a common scam that people used to run was to pay a ghostwriter ~$750 to 'write' a 'book' that had ~50-80 pages of bullshit word salad.

They'd then try to stick a real author's name on it, and try to sell it.

(Amazon has started putting in a little bit of effort at detecting this scam, so now the scammers have pivoted[1] to selling suckers a training course on how to do this.)

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


Also low-quality traditionally published garbage.


This is the equivalent of chiming in “75% of users eh? So that’s what… 3 out of every 4?”

Yeah, we all know that the problems that AI is likely to amplify are already problems. The question is whether or not those problems will get worse.

No matter how imperfect the current world is, one which is worse is 1) always possible and 2) worth preventing.


I was reacting to the implication that "traditionally published" is some sort of badge of quality.

Remember "Snooki"?

Traditionally-published.

I could name dozens, nay, hundreds of others.

In fact, "traditionally published" simply means that a big conglomerate thought it would make a buck. Quality doesn't enter into the equation. At all.


A big conglomerate [with access to market data on the demand side and tons of book optionality on the supply side] thinking it would make a buck absolutely is an additional filter for quality that doesn’t exist for AI-written self-published books.

Real businesses and real authors also tend to be concerned about getting sued for things like libel or defamation, or just generally having their brand reputations dragged through the mud after releasing mislabeled and subpar products.

It’s definitely an imperfect filter, sure, but then just see my original comment.


We could have a discussion about which celebrities you don't like but the issue here is material which is factually wrong and hence dangerous.


Like, say, astrology? Should that be banned?

No one put you, or anyone, in charge of deciding which information is "dangerous".


[flagged]


It’s the contrarianism fetish run amok


Yeah, I'm not sure what's up with these weirdos. It's like they expect to be taken seriously when their worldview is apparently based on absurdist philosophy.


It's taking ideas of cultural relativism to their most absurd, and useless conclusion.


Believing that adult human beings shouldn't be restricted from writing, or reading, anything they damned well pleased is not an "absurdist philosophy".


Thinking that random HN commenters should not be allowed to dictate what may, and may not, be published is neither "contrarianism" nor a "fetish".


> Remember “Snooki”?

> Traditionally-published.

Snooki was an established celebrity with a brand well before she was published at all; as far as I know her books were pretty much exactly in line with what people buying on the strength of their marketing would expect.

It may be “garbage” from the perspective of people with your or my tastes, but they were almost certainly not garbage from the perspective of the people to whom they were marketed. They weren’t, unlike the algorithmic (and previously, low-effort human written) spam that Amazon’s no-standards publishing and retailing policies effectively promote.


I think I'm at the point where "platform" stores are more hassle than they're worth. Saving a buck doesn't mean much when they aren't delivering the product you wanted.

Amazon is a huge vending machine stocked by anonymous third parties. That's useful in a few ways, and a huge mess in too many others. I'd already dropped them because their logistics are utter crap for me, but they seem to have no interest in cultivating consumer trust, so I doubt this will change after I move.


My intuition is that given the literal life or death nature of the enterprise, mushroom pickers should avoid leaning solely on any book, AI produced or not.

Mushrooms in different parts of the world look different. There can be subtleties that the books leave out. Anybody can publish a book and most books disclaim liability.

The best way is probably to find a local trusted, experienced mushroom picker that you trust and follow them around for a bit.

There is a reason that surgeons to be are not just given a textbook on surgery and told to just read the book and they will be fine. Instead they follow around experienced surgeons for a few years learning from them.


"My intuition is that given the literal life or death nature of the enterprise, mushroom pickers should avoid leaning solely on any book, AI produced or not."

Exactly. There is no second chance from death cap mushrooms, and in recent years they've spread to places where they weren't once found. Most know that some mushrooms can be dangerous but few know how truly deadly the amatoxin in death caps can be. People where I am in Australia have died only in the last few weeks from eating them despite warnings after previous deaths. Note the warning sign in the link below:

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

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


There's a reason for that and one of the reaaons is surgeries don't primarily happen a few weeks a year and require you to find happy amateurs to teach you around where you plan to do them going forward.


AI would be used (unsafely) to classify mushrooms, this exact prediction at 17:09 in Angela Collier's excellent "AI does not exist but it will ruin everything anyway" video, which I recommend for anyone interested in the subject: https://youtu.be/EUrOxh_0leE?t=1029


Hot takes aside, if you're looking for great foraging books, I'd suggest anything by David Arora, such as "All That the Rain Promises and More" and "Mushrooms Demystified". The National Audubon Society Field Guide to North American Mushrooms is, of course, pretty good too.

Outside of identification, "Chanterelle Dreams, Amanita Nightmares" and "Radical Mycology" are good reads too.


came here to suggest "All That the Rain Promises and More". it's a fantastic book with spore identification as well.

thanks to that book, I've eaten ~10 additional fungi that I would not have eaten before, and they were all quite delicious.

also, thanks to that book, I did spore tests on mushrooms that I was "pretty sure" were correct, and found out that I was wrong.

do your research, and you may find some delicious new meals. don't, and you may find death.


"...and you may find death."

Right, see my comment to jbandela1'


thankfully, I tend to stick with fairly easy to identify fungus, or even ones that have false versions that are not very toxic (false chanterelle are edible, but can give you a stomach ache, but still easy to tell the difference if you've been foraging before). I tend to stay away from morels, just out of fear of a false morel.

cauliflower, lion, an oyster, on the other hand, are very easy to recognize and the bolete family has some nice slimy cousins that are nice when dried and ground.

I think my favorite poisonous mushroom is what becomes the lobster mushroom due to an infection - great when infected, dangerous when not. thankfully it's easy to tell the difference.

I had one that surprised me, turned out to be quite delicious, but my book is put away, and I can't remember the name off the top of my head. now that it's coming to foraging season around here (pacific nw, usa), might be time to get it back out.


It’s only a matter of time before some AI ‘textbook’ on some obscure field of medicine or surgery gets somebody killed.


I've literally seen another commenter comment here on HN that chatgpt has given them more and better medical advice then their doctor. It's already happened probably and we don't know it. It's almost guaranteed that someone has asked for and taken bad medical advice from AI sources to their detriment.


It's only a matter of time before some internet "thing" on some obscure field of medicine or surgery gets somebody killed.

There is no Utopy. Embrace the nature of reality. AI will become norm, and it will bring all the good and bad with it.


We don't have to accept that.

We regulated formaldehyde out of food, mercury out of "medicine," and fire safety into workplaces. AI doesn't get to be special just because some business would rather be negligent.


"We don't have to accept that."

True if we took action, but look at what we've already accepted that's questionable. Just take how we've accepted how the internet has deteriorated and rotted over the past 20 or so years because we as a society have done nothing to stop the rot. Individually we don't like what's happened, collectively we're now powerless to do anything about it because it's too late.

The fact that Amazon has the irresponsible hide to sell books of this kind is proof enough. Amazon doesn't have to worry if someone dies, Bezos won't be charged. Things have changed.


it may be true that there is no other path, but we really can't help but notice that the bulk of human knowledge is being replaced by literal noise. that is going to be a real problem.


This feels like another place that libraries can be valuable, as curators of content.


If you're willing to spending a bit more on a book to support your local bookstore, please use something like https://uk.bookshop.org instead of shopping on Amazon.

I originally heard about Bookshop right here on HN years ago and have been using them since. It's a network of mostly independent bookstores that don't offer print-on-demand service which means you're less likely to get scammed by AI-generated spam.


If what that Twitter post is saying is true, I don't think this will last long. This sounds like it could be both civilly and criminally negligent if someone is actually harmed and the authors can't come up with a very convincing "we tried to verify but were wrong" story.

If they're literally just ingesting existing mushrooms books into an AI and publishing the output without verification or review, I think it would be hard to defend against negligence claims.


On the rare occasions when I order books from amazon, they are usually books that I've already read or borrowed from elsewhere, that I need for further reference or research.

Aside from AI-generated garbage, there are large numbers of works that are of just poor quality, with little to no editing, spelling mistakes, grammatical mistakes, formatting and layout problems, and (on one occasion) very smelly paper.

AI-generated rubbish is only the latest incarnation of this problem. I recall many years ago, purchasing a book on health and fitness that had many 4- and 5-star reviews. It was not cheap. It turned out to be a 50-60 page booklet that looked like someone had taken a bunch of webpages and clicked "print to PDF" on the browser - absolutely no proofreading or formatting. There were pages with the old IE missing image icon in place of what should have been photographs.

There is no reliable curation, a reputation scheme, or any trustworthy process on amazon to discover new works or explore new topics that return genuine material. This also extends to other goods such as household items, utility items, foods, and so on. It's scary that amazon also sells medicines - what's the guarantee that they are not fake, adulterated or toxic?

Amazon is a cesspool.


I’m looking forward to more B&N stores coming back to the Bay Area (https://www.smdailyjournal.com/news/local/barnes-and-noble-s...).

I used to love spending hours at Borders in Stonestown or the B&N in Colma. I had temporary solace with the Amazon brick and mortar in WC, then a B&N opened in Concord. The Emeryville B&N gives off a weird vibe for some reason.

My point is physical brick and mortar stores should be everyone’s source of truth for legit non-AI garbage book buying. Because of limited shelf space one is going to assume the books being sold have sort of vetting of sales and demand behind them. Vs Amazon having unlimited shelf space and letting garbage populate the virtual aisles.


I feel like at some point we should be holding Amazon responsible for deliberately creating a situation where it obfuscates culpability as much as possible for the fake quality of its products.

We crossed that line a while ago in my mind, when Amazon glibly allowed the Chinese junk product market, which it sends ambassadors to instruct on how to best ~~exploit its product, users~~ utilize it services to send dangerous products in the past.

"But it's impossible to police every product that's listed on the storefront!"

- Yeah, it's almost like this was a bad idea from the outset and shouldn't have been allowed to get this far in the first place.


- "should be holding Amazon responsible for deliberately creating a situation where it obfuscates culpability"

I'll be the unpopular downvoted guy and propose that it's actually a *dumb idea* to regulate the buying and selling of books, and to invent new forms of liability like "the bookstore didn't "vet" the books they were selling and therefore they're legally liable for the various consequences of people reading their wares". I think that's reactionary, illiberal, and destructive: and I think it will misfire. I don't want bookstores to empty their shelves of anything remotely "is this maybe theoretically dangerous in any context some rich person might sue us". I don't want a world of empty bookshelves; I don't want a censored, bowdlerized world where books are "banned by default" until they're proven "safe". I rather a vibrant, fearless marketplace of unregulated books, many of which *are* shit—but that's always how it was. This new, temporary emergency is neither.

We live, and we should live, in a culture that's fearless to speak and write and publish and bookmonger, and it's really dangerous to attack that foundation of liberal society which we depend on, just to go after one bad guy that's the villain of the moment.

("And if you cut them down, and you're just the man to do it, do you really think you could stand upright in the winds that would blow then?")


In most circumstances, I would agree with "that's reactionary, illiberal, and destructive".

And also, "I think it will misfire", I agree with this, too, the more I think about it.

However.

Amazon has deliberately manipulated this situation specifically to hide its own culpability behind the fear of such a thing backfiring. Amazon knowingly provides access to business tools to entities Amazon has a wealth of available information to indicate are not trustworthy. Amazon educates these untrustworthy individuals on how to apply for bogus/weird patent and copyright names so they can continue to shovel their dangerous products to unsuspecting consumers (Yes, an AI-generated mushroom harvesting guide is a dangerous, life-threatening product, there is no argument otherwise considering the state of AI information integrity). Amazon knowingly provides these tools to individuals who have already had products taken down previously for them being dangerous to consumers. It's even WORSE for the e-publishing side of things right now.

Perhaps my initial response was reactionary, but I can't help but feel Amazon should suffer for arranging this in some way.


> I feel like at some point we should be holding Amazon responsible for deliberately creating a situation where it obfuscates culpability as much as possible for the fake quality of its products.

You mean, like traditional chain-of-commerce liability for defective products?

AFAIK, despite trying repeatedly, in different jurisdictions, to use the “we’re just a marketplace where buyers meet sellers, not a real retailer” defense, they’ve pretty consistently been subjected to exactly that kind of liability when the issue has arisen, at least in US jurisdictions.


It doesn’t matter the subject. Know what you are reading and who wrote it. Be selective. If you want to learn about an empirical subject such as this, read things written by a human who is an expert in the field.

Don’t read Wikipedia, crap produced by “AI” programs, or anonymous semiliterate trash on the internet. It’s up to you to value your time and to filter what comes into your brain.


My advice that urge all to follow: Skip Amazon services and products entirely and almost anything else related to it or Bezos.


Turns out libraries will be sitting on gold mines, the only cache of books guaranteed to not be tainted by AI! :D


Holy shit, that's cold.


This is the classic tragedy of the commons.

Pro tip to Bezos: it has a tragic ending.


Life, only the 'selected works' procreate.


Why has the original title been edited to make it much less clear?


Because it was too long for HN. I had to work on it quite a bit to make it short enough. Sorry that I needed to manhandle it, I did not enjoy it either.


AI generated misinformation needs to be regulated in goods sold on tech platforms period. This is going to harm people


Brick & mortar bookstores about to make a huge comeback.


> The titles include “Wild Mushroom Cookbook: form [sic] forest to gourmet plate, a complete guide to wild mushroom cookery”

Does not sound like what an AI would write. AIs don't make obvious spelling mistakes like that. It is actually difficult for them to have bad spelling. Advanced LLMs like ChatGPT have very good writing, better than most humans, it is only when you look at the big picture that you notice something is wrong.

Here is an output from ChatGPT for illustration:

"As we delve deeper into the forest, remember that mushroom foraging is both art and science. Identifying hidden treasures like chanterelles or morels showcases your expertise. But with this thrill comes responsibility—ethics and sustainability are key. Leave no trace, as each mushroom affects the forest's balance. We also explore culinary creativity, from sautéed porcini to wild mushroom risotto. Let's continue, respecting nature's gifts and the wilderness."

To me, technically, that's good writing, and respecting wilderness is obviously the right thing to do. There is nothing "bad" about this. It is just that a book filled with pages and pages of such prose is utterly useless.


> Does not sound like what an AI would write. AIs don’t make obvious spelling mistakes like that.

Token-based LLMs like ChatGPT don’t. Character-based language models probably do make spelling errors.

I don’t know why we expect that scammers seeking to minize costs to produce, at the lowest cost possible, things that superficially resemble human-written books without much concern that they pass even the most cursory review of the contents would be consistently using best-in-class commercial models for that purpose.


The human probably wrote the title and the spelling mistake.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: