Should AI Write books?

[r16]

ahha , if we are caught with both hands in the cookie jar , it is a joke .

 

[Synobun]

These chatbots are effectively plagiarists, but of a sort that our copyright laws are ill equipped to deal with. All they do is scour the web for material on whatever prompt you give them. And they take whatever they find that best fulfills your request, including whole passages from authors who have posted material to the internet. They do that without attribution, and so it is plagiarism by any existing understanding of that concept. They often create a "works cited" page, but it itself is generally bogus, and never actually corresponds to the place from which they took the material.

I take the tone of the article Bird posted, though, to be a little tongue-in-cheek. The set of instructions it gives actually are a set of procedures that anyone who wanted to could use, but all of those resources are well known, as well, so if anybody wanted to do what is described here, they could already do it without the article. I say "tongue-in-cheek" because it starts from our premise that book-writing is difficult, and says, "but now, with all of these resources, it isn't any more. Ta da!" But let's imagine some would-be "authors" taking it seriously. It tells you first to go research the latest hot topic in publishing. Well, all people following the instructions would get the same result on that query. So let's say it's "how to lose weight on a vegan diet." We'd get a thousand chat-books on that topic. Moreover, Chat GPT would generate pretty much the same book for each of our thousand would-be authors. Then, the instructions for hiding that the book was AI generated involve swapping in synonyms, and we know how well that reads.

In other words, I think the results of the instructions would be absurdities: 1000 nearly identical books on how to lose weight on a vegan diet, with awkward synonym swaps. And I think that the author expects us to realize that, so the instructions aren't being offered in earnest. So that the point it ends up making is that, no, you actually need a human author to generate the new, fresh material that readers would really want to read.

I envy your optimism, but no, people genuinely suggest this process as a means of subverting the industry and making a quick buck. It is significantly impacting my industry. Flooding the market with AI-generated content makes it impossible to find the "new, fresh material that readers would really want to read." Algorithmic visibility is designed to be taken advantage of by AI, since it all uses the same base moving parts.

AI models lack a soul. They aren't real intelligence. They are trained off of material written by others, and are designed to simply reshape that material into different formats. If you tell ChatGPT to write a book, it is going to do so by stealing. It is incapable of creating original sentiment. The easiest way to realize this is to look at what happens to the AI models that are permitted to train off of their own generated material. It quickly becomes incoherent.

There is also the bigger question of how involved AI should be, fundamentally, with things that are the basis of human expression and joy. AI is a tool of productivity, not a tool of expression. Destroying the creative world and making it the realm of instantaneously generated drivel is genuinely dystopic. Nobody should support it. Nobody should go "Well, it's inevitable!" like it's the imminent heat death of the universe.

Speaking of AI replacing humans...

https://www.cbc.ca/news/canada/ottawa/artificial-intelligence-microsoft-travel-ottawa-food-bank-1.6940356

Microsoft has removed an article that advised tourists to visit the "beautiful" Ottawa Food Bank on an empty stomach, after facing ridicule about the company's reliance on artificial intelligence for news.

Published last week and titled "Headed to Ottawa? Here's what you shouldn't miss!" the article listed 15 must-see attractions for visitors to the capital.

The list was rife with errors. It featured a photo of the Rideau River in an entry about the Rideau Canal, and a photo of the Rideau Canal in an entry about Parc Omega near Montebello, Que. It advised tourists to enjoy the pristine grass of "Parliament Hills."

But the Ottawa Food Bank entry earned the most mockery in technology publications and on social media. The article called the food bank one of Ottawa's "beautiful attractions," before putting it third on the list.

Most of the entry simply describes what the food bank does, but it closes with a bizarre recommendation:

"Life is already difficult enough. Consider going into it on an empty stomach."

That appears to be an out-of-context rewrite of a paragraph on the food bank's website. "Life is challenging enough," it says. "Imagine facing it on an empty stomach."

 

[r16]

it is a legend that one of these bots was this nice charming personality which somehow went to how the Jews deserved to be killed , in the way of Adolf Hitler . Within 24 hours . As we will soon be assured that it was an early model and so on .

 

[Samson]

WRT the AI authors, someone has done it with the Maui fire, and it is quite successful. Akso some has also generated a book about that book.

'AI-written history' of Maui wildfire becomes Amazon bestseller, fuels conspiracies

A book that purports to recount the history of this month's deadly Maui wildfire has become a bestseller on Amazon, despite reviewers panning the work because its prose is on a par with that of AI.

The 44-page tome, snappily titled "Fire and Fury: The Story of the 2023 Maui Fire and its Implications for Climate Change," is a self-published work attributed to a “Dr Miles Stones” whose author biography states: "I'd rather not say." At the time of writing, Amazon US rated it the number-one bestseller in its environmental science category. It currently has 21 reviews on Amazon, all one-star.

Signs that the book was perhaps not crafted by a human include its rather repetitive writing in a dull style. Five of the seven sentences in the paragraph describing its content open with the same words, "The book," while the other two start with the book's title. Not exactly an imaginative start.

Another oddity is that Amazon lists Fire and Fury as having been published on August 10, yet the description of the work states that it "chronicles the events of August 8-11, 2023, when a massive fire swept across the island of Maui."

The good doctor’s author page lists other books he’s penned, one of which was published on August 11 and another on August 14.

Where does he find the time? Not even Reg scribes are that fast!

The timing discrepancy is, sadly, no laughing matter, as it is fueling bonkers conspiracy theories about who or what started the blaze, with some claiming the tome is evidence the disaster was planned or foreseen. For those who missed the news, at least 110 people were killed and the town of Lahaina destroyed after fires tore through the Hawaiian island of Maui starting August 8.

Fact-checking website Snopes has debunked the idea the book was published as some kind of eerie harbinger before the flames could engulf homes and businesses.

"The rumor about the book supposedly predicting the disaster was possibly prompted by the fact that its description on Amazon said the text covered 'the events of August 8-11, 2023,' while the publication date was August 10. It was unknown why the description included time after the book's publication date," Snopes noted.

One reviewer assessed the title as "inaccurate and insensitive… smells of AI, even before the smoke clears (literally)." Another described it as "a sloppily put together fake book," and claimed it was "generated by ChatGPT artificial 'intelligence'," adding: "Nothing of any value is contained within these pages."

Bizarrely, a book about the book has a 2.5-star rating, at time of writing.

That 14-page work is titled "Summary of Fire and Fury: The Story of the 2023 Maui Fire and its Implications for Climate Change by Dr Miles Stones EXPOSED.”

This one is also a bestseller, this time in the 30-minute education & reference short reads category. It’s even been included as a free read in Kindle Unlimited, Amazon’s subscription reading service.

Amazon has recently come under fire for allowing sales of books that seem to be AI-generated and claim to be authoritative but upon closer reading are quickly found to be slapdash and inaccurate. Jane Friedman, an author and reporter covering the publishing industry, found several books published under her name despite never having written them. They all have generic-sounding titles, and she believes they were produced by machine learning software.

Friedman criticized Amazon publicly, after which the e-commerce giant removed those suspect titles from its digital bookstores.

AI-generated books aren't strictly prohibited within the Amazonian digital tat bazaar, and it is not hard to find numerous publications on the site covering topics including travel and cooking.

 

[Synobun]

"The rumor about the book supposedly predicting the disaster was possibly prompted by the fact that its description on Amazon said the text covered 'the events of August 8-11, 2023,' while the publication date was August 10. It was unknown why the description included time after the book's publication date," Snopes noted.

Oh, that's a simple explanation. You can update everything about a book, even its content, after publication. Approval of these changes is managed by bots.

 

[Valka D'Ur]

You are describing plagiarism. It should be noted that plagiarism is not inherently a crime, but looked down upon in most circles.

Plagiarism is a crime in some jurisdictions. It's intellectual property theft, fraud, and can destroy the reputations of legitimate authors and artists whose work was stolen/whose names were used without permission.

Even the perception of plagiarism can destroy reputations. This also happens in fanfiction circles, btw. Some people have had their stories stolen, and of course they have fewer options to deal with it. As for me... I freely acknowledge which universes I'm playing in and which parts of them are from the source material and which parts I've created myself. And when I ran part of my story past MaryKB for some feedback and she said part of it reminded her of Game of Thrones - which I have never watched, and haven't read any of the books that show is based on - I told her that it's a coincidence, and right then realized that if I want this story taken seriously in the fanfiction community, Game of Thrones is something I have to avoid lest any part of it influence me even subconsciously.*

*David Gerrold had a situation like this when he wrote his "Trouble With Tribbles" script for Star Trek in 1967; there was a legal kerfuffle with Robert Heinlein, as parts of the "Tribbles" script were awfully close to the Martian flatcat subplot of Heinlein's novel The Rolling Stones. Gerrold acknowledged that he'd read the novel years earlier and was upfront about it probably having subconsciously influencing him as he was writing his script. Heinlein received some money for it, production of the script went forward, and the situation was resolved.

These chatbots are effectively plagiarists, but of a sort that our copyright laws are ill equipped to deal with. All they do is scour the web for material on whatever prompt you give them. And they take whatever they find that best fulfills your request, including whole passages from authors who have posted material to the internet. They do that without attribution, and so it is plagiarism by any existing understanding of that concept. They often create a "works cited" page, but it itself is generally bogus, and never actually corresponds to the place from which they took the material.

I take the tone of the article Bird posted, though, to be a little tongue-in-cheek. The set of instructions it gives actually are a set of procedures that anyone who wanted to could use, but all of those resources are well known, as well, so if anybody wanted to do what is described here, they could already do it without the article. I say "tongue-in-cheek" because it starts from our premise that book-writing is difficult, and says, "but now, with all of these resources, it isn't any more. Ta da!" But let's imagine some would-be "authors" taking it seriously. It tells you first to go research the latest hot topic in publishing. Well, all people following the instructions would get the same result on that query. So let's say it's "how to lose weight on a vegan diet." We'd get a thousand chat-books on that topic. Moreover, Chat GPT would generate pretty much the same book for each of our thousand would-be authors. Then, the instructions for hiding that the book was AI generated involve swapping in synonyms, and we know how well that reads.

In other words, I think the results of the instructions would be absurdities: 1000 nearly identical books on how to lose weight on a vegan diet, with awkward synonym swaps. And I think that the author expects us to realize that, so the instructions aren't being offered in earnest. So that the point it ends up making is that, no, you actually need a human author to generate the new, fresh material that readers would really want to read.

Absurdities or not, there are actually people who would buy this stuff and only find out later that it's fake. I don't know what Amazon's policies are regarding returning an e-book. I just know how frustrating it is to deal with some of the Marketplace sellers if the book they send isn't what I ordered, or if it's not in the condition specified. I also had to deal with a bait & switch situation where the seller contacted me after the sale and said he was in the U.S. and couldn't possibly afford to send it to me for the amount Amazon allowed him on the Canadian site, but he'd be happy to sell it to me on the American site - for a heftier shipping charge, of course. He was flabbergasted when I told him off and pointed out that it was his responsibility to read the rules and that there were plenty of other American booksellers in the Marketplace who managed to sell to Canadian customers without this sleazy tactic.

Customers have the option of posting an extremely low rating for customer service; ideally all Marketplace sellers want a 5-star rating, and get perturbed if they don't get it. Anything that's 3 or below can help to tank their reputation so fewer people would be inclined to buy from them. The bait-and-switch guy got a 1-star rating from me, and I didn't hold back with the reason why.

Purveyors of these fake books, on the other hand, aren't deterred by 1-star ratings. They just make or steal another identity and carry on. Amazon gets its cut, so prefers to do nothing unless someone threatens serious litigation.


As for the Maui fire in particular, there are climate change deniers on the Alberta premier's FB page who cite that as one of the wildfires that was deliberately set, either by government agents or "left-wing climate change hoaxers". These morons are convinced that every damn one of the wildfires in Canada - Alberta and the NWT in particular - were set by NDP supporters or supporters of Justin Trudeau, even though some of the fires started in areas so remote that no human could possibly have gotten there. Actual real helicopter pilots working in firefighting have stated that nobody could have set those fires deliberately and they were caused by lightning strikes.

Conspiracy propagandists eat up the fake information and gleefully spread it around social media - and since FB doesn't allow Canadian news links anymore (even Snopes got caught up in this insanity), debunking this BS is a lot harder.

So no, these fake books are not cute, they're not funny, they're not harmless. They're criminal.

 

[r16]

what conspiracy ? That they burned a town or whatever to clear land for development so that we too can buy villas to be neighbours with Bezos ? And covered it with a claim that it is the fault of global warming ? Already in tumblr . But , now you don't have fake your resources . You can have them for "real" !

 

[Takhisis]

I posted the article word for word without comment for folks to read. When comments showed up, I made the comment that it was going to happen. That is also a statement without an opinion about the goodness or badness of the facts at hand. For most of the articles I post, I do not comment on whether or not I agree or disagree (politics are an exception). My goal is not to sway folks with my opinion, just inform.

You are conveniently skipping the second post to which I am referring.
The future is not merely what's going to happen but what we're going to do. If you say that something is ‘the coming thing’ full stop then what you're saying is that we're faced with no alternative. History is deterministic. Ultimately, you seem to be arguing that resistance is futile.

 

[Samson]

Plagiarism is a crime in some jurisdictions. It's intellectual property theft, fraud, and can destroy the reputations of legitimate authors and artists whose work was stolen/whose names were used without permission.

Even the perception of plagiarism can destroy reputations. This also happens in fanfiction circles, btw. Some people have had their stories stolen, and of course they have fewer options to deal with it. As for me... I freely acknowledge which universes I'm playing in and which parts of them are from the source material and which parts I've created myself. And when I ran part of my story past MaryKB for some feedback and she said part of it reminded her of Game of Thrones - which I have never watched, and haven't read any of the books that show is based on - I told her that it's a coincidence, and right then realized that if I want this story taken seriously in the fanfiction community, Game of Thrones is something I have to avoid lest any part of it influence me even subconsciously.*

*David Gerrold had a situation like this when he wrote his "Trouble With Tribbles" script for Star Trek in 1967; there was a legal kerfuffle with Robert Heinlein, as parts of the "Tribbles" script were awfully close to the Martian flatcat subplot of Heinlein's novel The Rolling Stones. Gerrold acknowledged that he'd read the novel years earlier and was upfront about it probably having subconsciously influencing him as he was writing his script. Heinlein received some money for it, production of the script went forward, and the situation was resolved.

Plagiarism is different from breach of copyright. What you are describing may be (suspected) breach of copyright, but is not plagiarism because their is not intentional misrepresentation of authorship. If the AI is the original author is this case it would not be breach of copyright, as there is not copyright on AI generated works, but it could be plagiarism ro pass it off as your own. Intentionally ripping off a copyright work would be both breach of copyright and plagiarism.

 

[Kyriakos]

Plagiarism is different from breach of copyright. What you are describing may be (suspected) breach of copyright, but is not plagiarism because their is not intentional misrepresentation of authorship. If the AI is the original author is this case it would not be breach of copyright, as there is not copyright on AI generated works, but it could be plagiarism ro pass it off as your own. Intentionally ripping off a copyright work would be both breach of copyright and plagiarism.

But is there any guarantee the program (chatgpt or other) doesn't lift passages from other works without informing you at all?
I'd assume this is the norm; I even doubt there is any care that it slightly alters those passages.
So it could still be unintentional copyright breech.

 

[Birdjaguar]

You are conveniently skipping the second post to which I am referring.
The future is not merely what's going to happen but what we're going to do. If you say that something is ‘the coming thing’ full stop then what you're saying is that we're faced with no alternative. History is deterministic. Ultimately, you seem to be arguing that resistance is futile.

At the individual level we each can choose to engage new stuff or not. We each can choose to struggle against change in our culture too. And there are some fights that are likely to be futile: the war on drugs; cell phones and their future incarnations; AI doing stuff that humans do; medical use of genetics; etc. Yes, I do think that AI is coming, full stop. Currently, I do not actively engage in it. I'll be dead or mostly irrelevant soon enough, and it is best left to those for whom it will be a big part of their future to engage in how it gets implemented.

 

[Valka D'Ur]

But is there any guarantee the program (chatgpt or other) doesn't lift passages from other works without informing you at all?
I'd assume this is the norm; I even doubt there is any care that it slightly alters those passages.
So it could still be unintentional copyright breech.

Why would a scammer inform you that you're being scammed, or that your intellectual property was just stolen?

So far, it's not some AI deciding on its own to do this. There's a human in the beginning who decides this is an easy way to make a quick buck or 20 (have a look at that Amazon listing for the Maui fire book - nearly $18 for a physical copy, and even the e-version isn't cheap). By the time the customer realizes it's a scam, good luck in trying to get a refund. The criminal, in the meantime is sitting back with a giggle and a smirk, making a new account and using the AI programs to make more fake books to scam more people.

Amazon gets its cut, so why should it care?


BTW, @Birdjaguar or whoever on staff wanders by next: This discussion has gone on for quite awhile for a "Random" thread. It should probably be its own thread.

 

[Birdjaguar]

Moderator Action: Split from Random thoughts.

 

[NinjaCow64]

The answer to the question is “no” plain and simple.

 

[amadeus]

I don’t think a computer can do anything but analyze inputs—it can’t be creative on its own.

I don’t think it can be stopped by something like legislative fiat, but I don‘t think it will be necessary once the public find out that it’s all hype.

If there is any value in it, it’s going to be incremental change and not revolutionary; is there any technology that has come out where we’ve said everything changed? The big one to go to is the internet, but the earliest iterations of this go back 40 years if you’re talking about home computers, maybe more if you include things like telex printers and even things like the radio.

I think all the hype is out there just to sucker people out of their money.

 

[Kyriakos]

If I was to publish an "AI-produced" book, I would definitely be open about it being by "ai", since that is a way for it to sell due to the current hype.
Of course the worst idea would be to not edit the result, but that too can be "excused" as in "used further prompts to refine".
The story will not be as good as it would be if you just wrote it yourself. But anyone who has written anything, knows perfectly well that it is easier to edit something than to write it from nothing, so it's not unrealistic (and it will save a lot of time) to just take the "ai" stories and refine them.
After all, if you all you care about is selling some copies, there is literally no downside (well, other than the aforementioned very possible unintended copyright infringement; can't have knowledge of all phraseology used being or not being in existent books).
But you can include that in a note.

 

[r16]

a first thing would have been the pruning of Hollywood people . Where apparently 5 or 6 people would be in some office or whatever for a couple of weeks . And brain storm for ideas . Looking at many TV or movie productions , the thing about brain might look strange , but anyhow . AI set to cut that out . Only one guy and whatever . There has already been some AI made movie . Except there were a lot of people fixing things . None of them were named . For a Western context , this means it will be underpaid teams of Indians for a while . Then Africans when the Indians become too expensive . Meaning a supposed concentration of Liberalism and whatever will be shutdown . Not that Hollywood has stopped Trump but whatever . Have you seen young people having trouble adding numbers or similar arithmetic stuff ? That's because they have their phones and whatever available to do it in the easy and dependable way . This , still in an American context , provides the way to long lasting dream of the elites over there . A fiction , an illusion ... Of some advanced Mankind , without weak links in the machine . 1960s were tough , when the cogs did not like their chances of being drafted to die in Vietnam while they could get laid easily enough if they had some dope . No need to be cogs , you know .

this is a further step into an easier 1984 . All trade marked computer software , less people to rebel . Even if Lord Vader would be the first to tell of what strange kinds of rebels the Hippies would prove to be . Like it is them Hippies responsible for these days , right ? Somebody will end up burning . And it would be cooler for reputations for the post-fire times that if people did not act as advertisement agencies .

 

[uppi]

The instructions include a technique to disguise the true author of the text, allowing you to claim authorship. Sounds a bit like plagiarism to me.

I disagree. In my opinion, the AI is a tool, not an author. So if someone used an AI to write a book, I would still consider them the author.

Ripping plagiarism-relevant chunks out of other peoples work is a problem, of course. But if it can be ensured that this does not happen, I do not see a problem in principle. If someone manages to get an AI to produce a magnificent story, I am all for it.

The real problem is that AI enables creating a heap of low-quality work that no-one really wants. The question is, how to we make efficient quality control (for all books, not just AI written books)?

The story will not be as good as it would be if you just wrote it yourself.

Yes, if you are a decent writer. But there are certainly people who cannot write as well as an AI.

 

[Kyriakos]

Yes, if you are a decent writer. But there are certainly people who cannot write as well as an AI.

Unless one uses very refined prompts, and also provides cgpt with an elaborate idea, it is next to impossible to get it to produce anything other than utter garbage.
As you can see, with the bad (very general and not leading to specific forms/themes) prompts I gave it ^^

Or, you know, maybe the typo killed it. Typos aren't as anodyne as one might think.

If anyone's wondering, yes, I have gotten cgpt to write a half-decent story. But it took time and thought on my part, and the story still isn't good.
If I ever try again, I'd first create a table of analogies and feed those to the program, before asking it to use masks for the analogies and their prompted relations and depth of relations. That might work. But obviously it would work for the same reason you can program a sprite to move differently if you wish to assign it different speed stats.

 

[Samson]

I disagree. In my opinion, the AI is a tool, not an author. So if someone used an AI to write a book, I would still consider them the author.

There is spectrum, but the important point about plagiarism is the dishonesty between the author and the reader. If this is presented as an authoritative treatise by someone who has earned a doctorate in something, but is actually just the result of a statistical algorithm for "what are the most likely words to make a book about the Maui fire" then there is some dishonesty going on, and so "a bit like plagiarism".

Ripping plagiarism-relevant chunks out of other peoples work is a problem, of course. But if it can be ensured that this does not happen, I do not see a problem in principle. If someone manages to get an AI to produce a magnificent story, I am all for it.

What is a "plagiarism-relevant chunk"? We probably will over the next year or two know a lot more about what makes up enough to be a count as a derivative work, and what is fair use in this area, but unless microsoft completely losses with the Github Copilot case then I think the creative output is definitely not going to be a copyright breach of its source. Unless a plagiarism-relevant chunk is the same as that, in which case is it different from copyright breach, then that does not help us anyway.

As far as the general question of how much we should allow AI to intrude into the market for creative work, I think it is likely to be unstoppable and we should benefit in aggregate, but I definitely see problems and I expect there will be many more that I do not see.

The real problem is that AI enables creating a heap of low-quality work that no-one really wants. The question is, how to we make efficient quality control (for all books, not just AI written books)?

The way it has always worked before is you have trusted reviewers. It seems it must be possible, between hard encryption ensuring text is produced by who it says it is produced by and the general tend for people to talk about cultural works online should allow the better works to float to the top of the pile.