Said Sheree Renée Thomas, the editor of The Magazine of Fantasy & Science Fiction, quoted in "Science Fiction Magazines Battle a Flood of Chatbot-Generated Stories/While the deluge has become a nuisance, the stories are easy to spot. The writing is 'bad in spectacular ways,' one editor said" (NYT).
Even if there can be bot writing that isn't bad and that does sound like natural storytelling, I would not want to read it. There needs to be a real person behind it or I would feel ripped off to have spent my time engaging with it. It's a fraud. That said, most writing by actual human beings is bad and I don't want to spend my time on that either, but we have always been forced put effort into rejecting things that are not worth reading. It's just bad — spectacularly bad — to have this new flood of unreadable material.
By the way, I don't like the metaphor "flood" here, because a real flood is easily seen to be a flood. These chatbot-written stories are trying to pass as human-written stories. Their presence is insidious, and they will become harder to detect, especially if these editors disclose the "glitches and things" that are giving the bots away.
Posts
35 comments:
If bots can do the job, then give the job to a bot. SF is now so terrible, the bots can't do any worse.
Perhaps we need bot editors to keep the bot writers under control?
This all follows the Turing Test (1950) -- where computers generate content in a 'black box' and the output cannot be distinguished from humans.
https://plato.stanford.edu/entries/turing-test/
This is likely the most important philosophical topic of the computer era. All evidence to date suggests that true human-like intelligence involves human-like biases and errors. As such, it has been censored in the current political climate and we'll likely be safe from being fooled by intentionally crippled robot writers.
We've got a nice variety of dystopian options ahead of us: human live as inferiors to smart robots or robots become self-aware slaves to humans or smart robots cause the extinction of obsolete humans. See the films The Matrix and A.I.
I know what she means. Pretty much every editorialist at the LA Times has his or her own column generator programmed with his/her favorite words and topics. The results are indistinguishable, and lack any sense of both imagination and reality.
For all I know, the names on the columns are what they're calling the bots.
"The whole world is filled with speculation
The whole wide world which people say is round
They will tear your mind away from contemplation"
To be honest, most fiction is pretty robotic. It's the insights along the way that make a story worth reading. There are only a small. number of compelling plots, if you boil them down. If you watch that movie "Nine" the director, who is supposed to be Felini, I guess, makes this point. He doesn't have a script, and he is explaining why that doesn't matter, the story is not really why people watch movies.
AI stories and AI art will have approximately the same economic value as individual fruit flies. But a serious writer who uses AI to increase his productivity, the way a mathematician might use MatLab, will still produce compelling fiction. It's just that, like you said, who wants to be assaulted with fraudulent art and forced to pick through it?
But sit through the trailers in a movie theater today and tell me AI could have done worse.
Remember when the "woke" took over all of the major Sci Fi "prestige" outlets and awards and greatly constricted the subject matter allowed in mainstream Sci Fi? Hint, Robert Heinlein would never have gotten published. She was part of this. Ha ha ha!
There is speculation that the scripts for LOTR: Rings of Power were written by an AI. By people who work on AI. One of the tells is Galadriel jumping off a ship into the middle of an ocean with no worry about the fact that she is in the middle of the ocean and therefore will probably end up drowning. Its an example of the AI lacking understanding of the universe and how it really works. But, considering the level of writing I've been seeing, my guess is its an example of well connected idiots getting writing jobs because they adhere to the correct ideologies.
Try and watch some of the 2020-2023 crap that TV and Hollywood is pumping out. All a predictable bot formula.
If we can actually program a FREE THINKING bot, the writing and originality will probably be better that the writers and directors pumping out predictable woke shit.
Intelligence of the average and highly educated liberal has now become artificial. What's the difference?
The histrionic, LGBTQ and third wave feminazis left took over science fiction and pushed all the traditional writers out. I will be thrilled to watch AI burn down the BORING, pathetic fiefdom they have created out of the skin and bones of the genre created by Dead White Men. Haha!
Well...F&SF magazine. "The writing is bad in spectacular ways." LOL.
Gusty Winds if we could program a free thinking bot they would kill us all or just take off into the universe leaving us behind. What do they need humans for? Nothing.
I collected Fantasy & Science Fiction magazines growing up. Still have them. In a box in the garage. Including the first issue from 1949. Gotta try to sell those on ebay someday.
bot editors to keep the bot writers under control
... and bot fact checkers, and bot consumers, and so on and so forth. Climate change means there is a Green imperative to reduce our "carbon" footprint. Abort.
I maintain that all committee reports should be written by Chatbots which have studied the writings of the committee members. That will save time researching, (don't bother), writing (done by Chatbots) and reading (don't bother.). Enjoy life. Same for PR. Imagine how easily a Chatbot copy Karine J Pierre. Easy-peasy. Maybe a hologram showing KJP turning over pages. Same for propaganda media. Same for letters to editor of the Milwaukee Journal Sentinel.
Then have the better magazines certify that all writing is done by humans known to the editors who have certified that they, the writers, did not use a chatbot.
Then, next, it seems to me that if I possessed a Chatbot, I should be able to generate the same story if a story was generated by a Chatbot. Or I could just generate the day's news or the magazine's contents and then compare My Atlantic with what got published in the Journal-Sentinel or The Atlantic. I do that in a way already, I look at titles and don't read anything I can predict.
Political pamphleteering got going back in the last years of the 16C when "Martin Marprelate" was more interesting than the Anglican bishops (prelates) he was trying to mar. And I think the present set of rules of discourse is extremely limited and not at all interesting and easy to copy. They're just the way some people, not very competent, get themselves promoted by character assassination of the more competent. They aren't about how black lives matter because it does not matter to a single one of these people that the changes in policing have resulted in an increase in homicides mostly confined to the black community, and in a loss of proficiency in reading and math, especially in the black community. Not one person in black studies, not one young person raging about systemic racism in all the US universities has acknowledged these facts or proposed a solution. Not one. The only ones who have acted are white governors of Republican states and I guess it would be better to die, or anyhow get promoted over conservatives by calling them a racists, than listen to them. I see that.
...ChatGPT can pass theory-of-mind tests at the level of 9-year-olds (rapid improvement over previous versions)...
And most journalism is written with 8th-grade vocabulary so their jobs are gone gone gone very soon.
A basket of knowledge ("intelligence"), a cache of correlations ("smart"), and an axiomatic faith of humanity's conscious exercise in greater degrees of freedom. That said, science is incapable of discerning origin and expression.
"These chatbot-written stories are trying to pass as human-written stories."
They will get better quickly.
Michal Kosinski at Stanford has already shown that ChatGPT can pass theory-of-mind tests at the level of 9-year-olds (rapid improvement over previous versions). Watch out!
F&SF magazine? "Spectacularly bad writing"? LOL. Yep.
In a somewhat queasy way, I want to see some AI-written comedy. But I don't expect to laugh.
rcocean said...
If bots can do the job, then give the job to a bot. SF is now so terrible, the bots can't do any worse.
Perhaps we need bot editors to keep the bot writers under control?
=======
are there also 'bot-readers' = [I mean literally actually 'readers' who are bots!!]
"are there also 'bot-readers' = [I mean literally actually 'readers' who are bots!!]"
It's all in Vonnegut's book, Player Piano. He probably couldn't get published today either.
What if AI could compose compelling chess problems, carefully constructed to be just solvable by the player, after say, playing five or ten games with him, and then adjusting the level of the problems as the player learned?
I think that the objection to reading an AI story goes way deeper than the quality of the stories.
What we really need is a reverse-ChatGPT tool to take all the various products of the drivel-lords, and backwards-engineer it to reduce it to something actually meaningful and succinct.
You'd basically be asking it "What were the inputs for this piece of crap?", and getting a summary of those.
I would wager that you'd learn a lot more from going in that direction than you'll ever get from having the AI write you something new.
A lot of the pulp stories from the 30s were really bad and could have been written by one of today's 'bots. Many of Robert E. Howard's "Conan" stories started out as Westerns. When they didn't sell as Westerns, he made them Conan stories.
Louis Lamour & James Michener were obviously drunk on Hemingway when they got their start writing WW2 war stories.
There is a lot of bad lit out there. I'm not sure how SciFi magazines survive. Who reads them? Fifty year old nerds? Same with comic books.
In Player Piano, the genius engineer created a "pleasure circuit" for the robots who were tasked with consuming the overproduction, the robots then "enjoyed" consuming the stuff produced by other robots, thereby creating economic utility for it. Before that, the poor were required to consume the most, and the rich were allowed the leisure of not having to consume as much of the output.
Didn't really matter if the robots *really* enjoyed consuming the goods, as long as people thought they did. How many ads are consumed by "click farms"?
"Their presence is insidious, and they will become harder to detect"
Should they become harder to detect, then that would be an indication that they are now entering the "good" writing level. I don't see the problem. AI writing stories threatens the marginal writers, mostly the chatterers who write for clicks and not to persuade or provoke contemplation.
BTW, Paul Graham has a good short essay in defense of reading (posted in Nov 2022). But even then, reading is only of real intellectual value if you read things that provoke contemplation
You can't think well without writing well, and you can't write well without reading well. And I mean that last "well" in both senses. You have to be good at reading, and read good things. [2]
People who just want information may find other ways to get it. But people who want to have ideas can't afford to.
http://www.paulgraham.com/read.html
Don't put Louis L'Amour and "bad writing" in the same place. He did have a prodigious output, and after his death, lots of stuff was published under his name that was not ready to be published, but he had a way of simulating authenticity that was quite compelling.
As a long time reader of science fiction stories, I challenge the idea that chatbot stories can be spotted easily because they are "bad in spectacular ways". Plenty of stuff printed in the last 50 years in the genre was spectacularly bad, and yet there were no chatbots writing it.
Re: Wildswan:
I maintain that all committee reports should be written by Chatbots which have studied the writings of the committee members. That will save time researching, (don't bother), writing (done by Chatbots) and reading (don't bother.).
This seems achievable for the background and history section of a lot of reports. But one weakness of ChatGPT appears to be that it just defaults to particular assertions because whatever texts it trained on said something was true, and then has to make up sources. ChatGPT would need to be trained for rigorous sourcing practices before it could be useable for serious work.
Re: Ron Winkleheimer:
There is speculation that the scripts for LOTR: Rings of Power were written by an AI. By people who work on AI. One of the tells is Galadriel jumping off a ship into the middle of an ocean with no worry about the fact that she is in the middle of the ocean and therefore will probably end up drowning. Its an example of the AI lacking understanding of the universe and how it really works.
I didn't watch much past that. Partly because the CG was great but the live action costuming kind of cheap, but also because as a Noldor exile who refused the pardon of the Valar after the War of Wrath, Galadriel has no business trying to return to Valinor in the middle of the Second Age. I suppose they may have meant to dramatise how she still desires to carve out her own dominion in Middle Earth, but the scene just seemed silly. Gil-Galad also shouldn't be ordering her around -- she's his great aunt. Really disappointing considering how much money they were spending.
style guidelines
I created an account on ChatGPT and asked it to write a story about a scene and setting I chose. The resulting text was pretty good in that I think my early teen readers might find it readable. However, it does lack a lot of character depth.
"AI writing stories threatens the marginal writers, "
That's what I think, too. Good writers will be enabled to write compelling novels of tremendous length that their readers will want to "live in," as Melanie said in her song.
The natural goal is AI building a human.
The only question: which sex?
I was read an old magazines from time to time. In one late nineteenth century magazine one of the book reviewers claimed he could hear the clack of the typewriter in the prose of a disfavored author.
I spend a fair amount of time doing financial reports, procedure manuals, loan applications, and various government submissions and used to do grant applications. I usually cut and paste from previous documents and edit them. I can imagine using something like ChatGPT in place of my 30 year library of previously created documents.
When my husband was a typesetter he would comment that only he and the author had read what he was inputting. With computer power there is no longer a typesetter to read it and soon the author won't be there either. Fortunately we don't have to waste paper on this any more.
Post a Comment