generatingfail

[generatingfail]

[generatingfail]

AI copyeditor is fail

An unidentified scholarly journal (whose editors should be turfed out for negligence) sent author Jill Walker Rettberg's article manuscript through an "AI copyeditor" called Paperpal. Result: total unutterable fail.

Middle schoolers know better than some of the shenanigans shown!

[generatingfail]

Siiiiiiiiiiiiiiiiiiiiiiiiigh.

ChatGTP rotting away your brain

[generatingfail]

One day when I was extra-angry about gen AI I started collecting metaphors for it.

Because the AI hypesters have brain and intelligence metaphors, and those are unfortunately super-powerful.

This is a good one and I'm adding it to my list.

really enjoying calling chatgpt a 'magic 8ball' because while far from a perfect analogy i feel like it is a very good deflationary weapon against both hype and antihype that helps bring things into perspective by detaching the actual thign from the cultural object it's attached to. "this magic 8ball is going to become sentient" i dont think thats true. "if you talk to a magic 8ball it will kill your soul" that one doesnt sound true also to be honest

[generatingfail]

This is terrific. If I were anywhere near Seattle, I'd be there!

ConCurrent Seattle, or, What Happens When I Get Cranky

I was going to do this whole post about how genAI is ruining so much of what makes us human - human thought, human creativity, human connection - but as it so happens, I am also tired. So let's run this down:

On April 30, the con chair of WorldCon posted a statement answering questions about whether WorldCon had used ChatGPT in any of its processes. The answer was yes. It was used in panelist vetting.

The reaction was immediate and negative. Writers in general don't like genAI, especially writers who've had their work stolen for the LLM to train on (which is most of us). We also dislike the environmental impact of this!

Because it was used for vetting, a lot of panelists chosen for WorldCon began doubting whether they were chosen fairly.

On May 2nd, an apology was issued. The chair promised another update on what actions WorldCon would take to remedy the problem. Unfortunately, WorldCon had already lost a lot of trust in the community.

An explanation of what ChatGPT was used for and the prompt given to it dropped on May 6.

I was one of the authors who'd applied to WorldCon but got ghosted (I don't even get newsletters from them despite signing up). I'm not on the panelist list. Neither were some authors whose voices I thought truly belonged at WorldCon. Well, when I get cranky, I move. I decided that I was going to create a fringe con to run during one day of WorldCon, with stellar programming and wonderful panelists - and I'd do it with no genAI. No LLMs, no GPT, nothing but human brains and human work.

I present to you: ConCurrent Seattle.

Credit goes to @inkcurlsandknives for the logo design!

ConCurrent is a one-day, in-person only, masks required SFFH programming track intended to be an alternate to WorldCon. If you're in Seattle on Thursday, August 14, 2025, please drop in to see us! Admission is free, but registration is required (we'll have that process sorted out soon). I've got a bananas lineup of authors, I can promise you that. The sparklecorns will be sparkling.

But to make it run, ConCurrent will cost us 5000 USD. I've been crowdfunding for the last couple of weeks and we're halfway there! That's the good news. So now we need another push to get fully funded. If you wanna put your middle fingers up at genAI, I'd super appreciate it if you could send some dollars my way.

Yeah, that's it. This con is happening. It's gonna be a blast. I'm hyped for these panels. Please help defray the cost and I'll see you in Seattle!

(And after you go to ConCurrent, you can go see @charliejaneanders with Annalee Newitz, Cecilia Tan, Becky Chambers, Darcie Little Badger, and Andrea Hairston at Writers with Drinks at 7 p.m. at Town Hall Seattle!)

[generatingfail]

As an educator, I am taking this very seriously indeed.

Okay parents and teachers and guardians of children, your priority mission right now is to defend them against GenAI.

Set an example, don't use it yourself, call it out when you see it, don't accept it in daily life, don't treat it as fun, impressive, cool, pretty.

Why?

GenAI is being sold to us by trying to convince us that we shouldn't bother trying to create things ourselves because a machine will always be better at it.

You do not fucking want kids to grow up believing that they aren't good enough to create stuff. You do not want them to get used to the idea of needing a machine to do their thinking for them.

You do not want them giving up on their own ideas because they have been conditioned to believe genAI always gets better and more valued results.

You absolutely do not want them TRUSTING an LLM's 'opinions' 'analysis' or 'research' over their own!!!

Kids are especially vulnerable to having the development of their creative thinking be stunted by being given no time and space to work things out for themselves. Kids are impressionable and impatient. Kids are dependent and inexperienced. Kids are defenceless against the impossible numbers of addictive low-effort instant-reward products and apps the world is pushing on them.

GenAI right now is not a fun harmless toy, it is a scam that will make your kids give up their confidence and potential for a quick shot of dopamine. Which is what its investors want, obviously- convenience and assistance are not the goal, and probably never were.

[generatingfail]

holy incisive metaphors Batman

may I quote this? like, professionally? and if I may, how would you like to be credited?

there are already quite a few articles that talk about people starting to believe chatgpt is accessing like. a higher form of power and secrets to the universe and its actually really scary to see how quickly people can fall for it and start believing it without any questions. i can send you some archive links if you'd like but basically yeah. it's about what you can probably imagine.

The human brain is a pattern-seeking lab rat that gets a hot bump of cocaine every time it connects two dots and we should not be evaluating the value of those connections based on the enthusiasm of the rat

[generatingfail]

Deep Blue is 30 years old and was capable of defeating chess grand champions. It could be housed in a single cabinet.

ChatGPT spans untold data centers devouring massive amounts of electricity and it got its ass whipped by an 8 bit gaming console from the 1970s.

[generatingfail]

"AI is gonna take your job!" Not when it gets the very simplest parts of your job right less than 60% of the time. (The subhed rounds up. The actual percentage was 58%.)

Also not when they blow confidentiality high, wide, and handsome.

Source.

[generatingfail]

Do not trust your life to AI fail

So, they tested people-with-web against people-with-chatbot (various) in the not-a-game of diagnosing medical issues, and as you'd guess, the chatbots failed people.

People-with-chatbot got it right just over one-third of the time. People-with-web got it right about half the time.

This is obviously a reason to talk to medical professionals instead of self-diagnosing!

But it's also another delightful demonstration of AI fail.

Source.

[generatingfail]

"AI Engorgement" refers to the phenomenon where an AI model absorbs too much misinformation in its training data. This corrupts the model's base truth, leading to strange glitches. An engorged image model, when prompted to create images for "cat" and "historical", generated the following:

AI Engorgement may trigger a complete dissolution of truth within a model, leading it to back conspiracy theories, clearly debunked facts, and mistake fiction for reality. It is believed that the systematic siphoning of unreality by AI datasets is already revealing signs of engorgement in every major model.

[generatingfail]

Source.

This has been happening to my reference-librarian friends for at least a year and a half now. Guess the bullshit is spreading to collection developers.

If folks don't know that the bullshit machines make up bullshit, I don't know what to say at this point.

[generatingfail]

AI IS FAIL. It's particularly insulting to librarians, archivists, and other information professionals.

The thing I wasn’t totally prepared for with a workplace trying force AI crap on me, the corporate librarian and archivist, is how insulting it is.

Every “feature” they are trying to sell me on I can do better and faster using methods I already have available.

My (extremely valid) concerns and arguments about why I don’t think doing these things with AI is actually a good idea are being countered with “well everyone is doing it/following policy” which is fucking gross.

My concern that everything that has been proposed actively makes more work for me, not less as they are saying, but also makes my job worse, harder, and more demanding.

Anyway. The Luddites were right and I need a replica Enoch’s hammer for my desk. (And everyone should read ‘Blood in the Machine: the origins of the rebellion against big tech’)

[generatingfail]

Would anyone like popcorn?

May the Middies be chewed to death by the Mouse.

ETA: a little more detail--

Plaintiffs to the lawsuit include Disney Enterprises, Inc., Marvel Characters, Inc., MVL Film Finance LLC, Lucasfilm Ltd. LLC, and Twentieth Century Fox Film Corporation (collectively “Disney”), as well as Universal City Studios Productions LLP and DreamWorks Animation L.L.C. (collectively “Universal”). The lawsuit was filed at the U.S. District Court of the Central District of California on June 11. Specifically, Plaintiffs seek relief over Midjourney’s alleged direct copyright infringement, and demand that Midjourney be barred from further alleged infringement and pay either actual damages (an amount of damage that can be proven Midjourney is responsible for) or statutory damages limited to $150,000 per infringed work. Furthermore, for secondary copyright infringement, they also demand actual damages or statutory damages, again, up to $150,000 per infringed work.

[generatingfail]

Excellent. Someone call TIME Magazine.

I think "slop" is a strong contender for world of the year 2025, as it seems to be universally recognizable as a description for the majority of output produced in the name of Return on Investment.

Sure it means algorithmically generative slop now, but look at housing construction, durable appliances, food, automobiles, you name it and tell me it's not just more slop. If enshittification is the process, slop is the result.

[generatingfail]

As someone who teaches a fair few CS majors... thank you. Please consider evangelizing your fellow majors!

(No, I don't teach CS except glancingly and to layfolk. CS majors have to take gen eds.)

Please forgive me for ranting, but...I am so tired of AI. Just so tired. I don't want Microsoft Copilot, or Google Gemini, or Meta AI, or whatever other energy-sucking, water-wasting, mediocrity-spewing LLM is currently being thrust upon me. I just want to be left alone to create in peace.

[generatingfail]

Agreed.

Please forgive me for ranting, but...I am so tired of AI. Just so tired. I don't want Microsoft Copilot, or Google Gemini, or Meta AI, or whatever other energy-sucking, water-wasting, mediocrity-spewing LLM is currently being thrust upon me. I just want to be left alone to create in peace.