Clarkesworld
This is a problem for short fiction submissions and it's not just going to go away. The link goes into details, but this is a graph of submission bans since 2019. Plagiarism and bot-written spam.
http://neil-clarke.com/a-concerning-trend/
A Concerning Trend – Neil Clarke

Feb 15, 2023 at 11:21pmWeb
Show threadBogi TakácsFeb 15, 2023
@clarkesworld 1. Absolutely stunning, and not in a good sense - I'm sorry to hear you are dealing with this 😲 2. If your site is hammered by Mastodon links, as it seems to be right now? try this https://jort.link/ (Had to do this for one of my part-time jobs, I kept on knocking the webstore offline!)
jort.link - a solution to fediverse request floods

A URL redirector and shield to solve fediverse request floods.

Show threadA. P. HowellFeb 15, 2023
@clarkesworld Thank you from shifting this from "seems like it'll be a problem" to "let me show you some data on this thing that is already a problem."
Show threadM EiloFeb 15, 2023

@clarkesworld I recommend using the free AI text tester GLTR. Especially if it can be APIed into the submission form directly so fails dont increase cloud costs.

http://gltr.io/dist/index.html

GLTR (glitter) v0.5

Show threadClarkesworldFeb 16, 2023
@BlinkPopShift That hasn't been updated in a while and misses a lot. Their own site says "It might not be helpful to detect texts for recent models (ChatGPT)." None of the detectors are particularly strong.
Show threadM EiloFeb 16, 2023
@clarkesworld Oh yup! Definitely. OpenAI announced a newer model but I haven't seen anyone get access yet. But starting with an older model might let you dev new processes new models could plug into?
Show threadSL HuangFeb 16, 2023

@BlinkPopShift @clarkesworld I would have large concerns with a magazine implementing an auto-reject of suspected AI subs :-/ So many tools are not sufficiently accurate + complete black boxes where biases can't be checked

Years from now we could easily find an auto-reject tool had unacceptable false positives correlating w newer writers/non-native English speakers/ppl who used dialect -- we've seen similar happen IRL elsewhere -- & we already have big issues with the field being inequitable...

Show threadClarkesworldFeb 16, 2023
@slhuang @BlinkPopShift This is just one of the reasons I wouldn't trust those programs with anything beyond human-reviewed scoring. I've seen this kind of mistake in detection, particularly in non-native English speakers. The tech is unreliable.
Show threadSL HuangFeb 16, 2023
@clarkesworld @BlinkPopShift Appreciate that, Neil (also what you said about sub fees / postal mail not being solutions). I'm sympathetic that this isn't an easy problem! esp as CW is one of the last pro markets always open / not privileging solicitations
Show threadBogi TakácsFeb 16, 2023

@slhuang @clarkesworld @BlinkPopShift

As a non-first-language speaker of English, I'm glad for that.

I've also seen people gathering data that prose by autistic people is more likely to produce false positives on these filters than prose by non-autistic people. So that's possibly a thing too.

Show threadSL HuangFeb 16, 2023

@bogiperson @clarkesworld @BlinkPopShift Oh, if research like that crosses your dashboard again, I'd be very interested in being tagged so I can come read it! only if convenient ofc :)

(I just tried doing a search and got some supremely awful results that I maybe should have expected but were NOT what I was looking for. They were....not about writing)

Show threadBogi TakácsFeb 16, 2023

@slhuang @clarkesworld @BlinkPopShift It was somewhere here on Mastodon, just informal polls, but quite many people responding.

(I will NOT try a search, thank you for the warning 😬 )

Show threadDr. Damien P. Williams, MagusFeb 22, 2023
@slhuang @BlinkPopShift @clarkesworld especially because most of the gpt checkers out there misidentify material from neurodivergent people as "likely generated by a.i."
Show threadHTTP 1.1/418 TeapotFeb 15, 2023
@clarkesworld @bhawthorne Huh! I had not heard this was a thing but it makes sense that it would show up in our current hellscape - only have to succeed a small percent of the time to make it a profitable scam.
Show threadJoan Combs DursoFeb 17, 2023
@clarkesworld @[email protected] @rmd1023 Wonder if those “authors” are later “publishing” the rejects on that big website that sells fiction to readers and lots if other stuff.
Show threadCarlos ArizaFeb 16, 2023
@clarkesworld I wonder if this may force a change in how you receive submissions? Maybe they need to come from referrals (which creates another hoop to jump) or maybe people need to pay a small fee to join a Patreon in order to submit? Even a small nominal payment should discourage spammers... maybe the solution is to go back to mailed printed submissions?
Show threadClarkesworldFeb 16, 2023
@carlos_ariza Paying to submit is a shortcut to being drummed out of the SFF community. It's just not acceptable behavior. If new writers end up being shut out from most opportunities, the field will eventually suffer dearly for it.
Show threadCarlos ArizaFeb 16, 2023
@clarkesworld it's good to hear that you want to encourage new voices, and it's concerning that it is becoming harder to get noticed as a new writer bc of the deluge of AI-written submissions :/
Show thread👩‍👩‍👧🐈🌁👩🏻‍💼🏳️‍🌈🤵🏻‍♀️💄👠❤️🤯💡👩🏻‍💻Feb 16, 2023
@clarkesworld I ponder the analogy with photography's impact on painting. As cameras became more common art explored surrealism. I can imagine a future in which my Kindle can compose space opera on the fly with exactly my preferred mix of aliens, mystery, and romance whenever I need escapism, and human-authored stories have gone all House of Leaves or Feersum Endjinn.
Show threadAlaricFeb 16, 2023
@myx @clarkesworld it could even write stories with you as the main protagonist, given a bit of your own personal data to learn from
Show threadMarc TriusFeb 16, 2023
@myx @clarkesworld to me this sounds awful 😬
Show threadAustin Beeman 🍷📷🛫🛸Feb 16, 2023
@myx @clarkesworld Or a world where human writers are paid to show up in the houses of the wealthy to write in-person stories that exist only as limited edition 1 of 1.
Show threadJohn Skiles SkinnerFeb 16, 2023
@clarkesworld this makes me oddly much more sad than I would have predicted 🙁
Show threadseth L.Feb 16, 2023
@clarkesworld Woah. That's... that's something else.
Show threadLizzy MillerFeb 16, 2023
@clarkesworld I yearn for the "good old days" where I could ban article spinner spambots who tried to submit the Wikipedia article about Dick Cheney. With even his first name replaced by a synonym.
Show threadAnn LeBlancFeb 16, 2023

@clarkesworld Huge sympathy to you and the slush readers having to deal with this surge.

This situation feels similar to my experience fighting online credit card fraud at a payment processor. While it's easy for humans to detect fraud, the rise of automated tools allows bad actors to overwhelm fraud detection systems with sheer numbers. It's a very non-trivial problem space unfortunately.

Show threadClarkesworldFeb 16, 2023
@RobotLeBlanc it's very similar and how I've been explaining to people why I don't see a solution. People with much more at stake have been fighting fraud for much longer and haven't "won" despite the resources they can throw at it
Show threadAnn LeBlancFeb 16, 2023

@clarkesworld Exactly. There's no magic bullet, and the bad actors are constantly evolving.

I'm happy to share what worked for us in managing the payments queue, if that's helpful. From reading your blog post, it sounds like you're already on the right track though.

Show threadBodhipaksaFeb 16, 2023
@clarkesworld @cliffjones Sorry you’re having to deal with this. It’s easier to create bullshit than it is to shovel it.
Show threadOceanicEternityFeb 16, 2023

@clarkesworld

It is sad to see the exponential growth of Bullshit.

Show threadDoubleMaulerFeb 16, 2023

@clarkesworld Is shadow banning an option? Maybe letting the cheaters submit, but internally throwing to the bin everything they send from specific users, IP addresses, etc.

If they don't have feedback, it would be less likely for them to change strategies, like creating new users.

Show threadClarkesworldFeb 16, 2023
@doublemauler Given what I know about this crop, shadow banning and honeypots wouldn't be very effective. I've used these techniques in other systems I've developed.
Show threadAnoninymityFeb 16, 2023
@clarkesworld @clarkesworld@mastodon. Google and Bing do not demote A.I. generated content. After all, with Bing Chat and GPT3 in the new Bing search, and Google racing to incorprate their language model into their products, they are not going to demote something that they themself introduced and are using. "Google doesn't care if the content is created BY people, as long as it's created FOR people."
Show threadpnwpeteyFeb 16, 2023
@clarkesworld Yet, after 30 years of trying, I still can’t get a story accepted to save my life. 😂🥺
Show threadDerek SoFeb 16, 2023
@clarkesworld Horrible!
Show threadDerek SoFeb 16, 2023
@clarkesworld I've heard of something called BrightID that people use to prove they are a unique person online (which might help stop fraudulent submitters from trying again), but I have no idea how it actually works or if it would be functional in this context.
Show threadClarkesworldFeb 16, 2023
@derekso Yes, I've heard of them mentioned, but don't know much about them. Site was unreachable when I tried.
Show threadEric PierceFeb 16, 2023
@clarkesworld I run a pop culture site and this matches my experience. The sheer quantity of drek is astonishing. And overwhelming. Our submission turnaround times have more than doubled.
Show threadBill SeitzFeb 16, 2023
@clarkesworld Publish them all (online) and let readers vote for the good ones.
Show threadPJ "Vote Green" CoffeyFeb 16, 2023

@clarkesworld

Well that is the sort of thing that people thinking about:

#AI #AIEthics #ChatGPT #AIJobs and #AIWriting

would probably be interested in. Thanks.

Show threadThe Duality of XanFeb 16, 2023
@clarkesworld This is going in exactly the direction I feared it would. Harmful to all of us, particularly those who aren’t well established yet.
Show threadAustin Beeman 🍷📷🛫🛸Feb 16, 2023
@clarkesworld What happens in 6 months if this trend continues?
Show threadClarkesworldFeb 16, 2023
@austinbeeman We don't know yet, but it won't be pretty. It's possible this plateaus, but we're not counting on it. Still have a few tricks up our sleeves.
Show threadJasonFeb 20, 2023
@clarkesworld this is bleak
Show threademil čolićFeb 21, 2023
@clarkesworld this is so grim
Show threadThe Febrile MuseFeb 22, 2023
@clarkesworld Go back to SASE.
Show threadStories and FolkloreFeb 22, 2023

@clarkesworld I just read about this today, it is absolutely bonkers.

I really hope, we ca a community can work out a sensible way to manage all this noise.It appears to be becoming crippling.

Show threadPeculiar SusanFeb 22, 2023
@clarkesworld As frustrating as this is for you, imagine how we emerging writers feel.
Show threadClarkesworldFeb 23, 2023
@Sjb127 We're not giving up on any of you.
Show threadPeculiar SusanFeb 23, 2023
@clarkesworld I know you’re between a rock and a hard place here.
Show threadGoldaApr 23, 2023
@clarkesworld well - we need trust graphs and credentialling, perhaps - at least friends attesting and staking their credibility on each other that they are not using bots