David Revoy | Mastodon | Bonus Timelapse
Transcript
Malicious Compliance
A comic in four panels:
Panel 1. The headmaster bursts through Cepper’s door, slamming it, furious. We see Cepper in the foreground, sitting at her desk, bored, not reacting.
Headmaster: “Cepper! I’ve heard you’re refusing to use your Avian Intelligence!”
Headmaster: “As a scholar of this university, you must! Your pay depends on it.”
Panel 2. A closer shot of the headmaster as he continues his tirade, gesturing dramatically but reading the scroll of a Avian Intelligence Parrot with a little hat and a beard with on it “HR”.
Headmaster: “We also need your Grimoire of Shadows and Fire by the end of the month!”
Headmaster: “So, use your AI! This is non-negotiable, our investors require it!”
Panel 3. Cepper turns in her chair to face the headmaster, revealing the AI Parrot on her desk. She offers a large grimoire to the headmaster (the book remains off-screen).
Cepper: “Calm down… I am using it, and it’s already done.”
Cepper: “The investors wanted high-volume output, right?”
Panel 4. The headmaster is now holding the book, and the audience can see the cover. The cover features Cepper’s face with a dramatic, clickbait-style background. The title is in bold: “Shadows and Fire: 100 Tips (Number 72 is a Game Changer!)”. There is a big arrow pointing to a fire with the word “HOW?!” in giant letters and also “WHAT’S NEXT?!”. The headmaster is surprised. Cepper She has a dark, malicious smile.
Cepper: “I’ve given them exactly that!”
Cepper: “Enjoy the 500 pages of untested slops!”
Headmaster: “?!”
You must log in or # to comment.
Actual conversation between my so and boss. That bloke uses ai for everything but gets annoyed my so tags it ai.
Ugh, this just reminded me that I had a system architecture document written up at work but needed a diagram to go along with it. One of our enterprise architects recommended that I just feed the document into their new AI diagramming tool.
I will say, it was very nice that it did the whole thing in less than a minute. I wasn’t a big fan how it was completely wrong. The only thing I can think of is that more people prefer pretty pictures done quickly rather than accurate pictures that take time.
that’s why ml should be used complementary
Pretty but inaccurate pictures done quickly sounds like management’s wet dream.
If it is any consolation, in my experience no one will ever read the system architecture document anyway. So if it’s totally wrong, no one will know.
You say that until you have to come back to the project after months and completely forgot what you were doing there. In that case having good system architecture documentation is a life saver.
Guess my career has gone such that I’ve always looked to code and git history to wrap my head around, because the documentation people write is unreliable.
Particularly in enterprise, where for some reason there’s some “developer” who writes a word document and then buggers off while someone else actually does the coding and the relationship between actual code and document drift because the “architect” never had to reconcile his document with the realities.
If that’s the case looking at the code (depending in the size of the codebase and how bad the differences are) might be the better call. The only thing worse then missing documentation is wrong documentation.
Yeah, my beef was that I had to present that diagram to a review board today. So ain’t no way was I gonna present slop that didn’t make any sense.
that brings me back to the time i was phd student and my supervisor forced me to publish something that i knew to be wrong. the paper got published… but my name disappeared from it ;)
Except, why tarnish your own reputation? For true malicious compliance, you need to publish it not under your own name, but rather under the name of avian intelligence, and call it “100 ai-generated and completely untested slop tips on shadows and fire, compiled under the express direction of <Professor’s Name>”. The only part of it that shouldn’t be AI-generated is the introduction and acknowledgements section, in which you publish every record of them demanding that you use this piece of shit, including each of your strenuous warnings.
In the acknowledgements section, point out five different times where the rituals will, if completed to the AI’s specifications, cause the imminent demise of the reader. Thank AI and the professor who foisted it upon you for these “contributions”. Offer no solutions, but suggest that volume 2 of the series will feature precisely zero avian intelligence, and will be written by you. Then, publish.
ETA: Source: Am chemistry teacher. Was handed ai-generated curriculum. Pointed out multiple different times when the curriculum could lead to the potential death of students or people in their care (as a representative sample, the curriculum said that one of the most common IV normal saline solution ingredients was KCl —the lethal ingredient in lethal injections— and there was a lab which called for students to boil 95% isopropanol on a hot plate at max heat —550C— outside of a fume hood, with no procedure or safety section). Am no longer expected to use AI slop curriculum. You have to make it clear that using this shit puts the institution into active legal jeopardy.
I understand that this is Off-topic but what does ETA mean? I can’t help but read it as “Estimated Time of Arrival” or possibly “Elvis Tribute Artist”, neither of which seem right. Please excuse my ignorance and help me learn. 😊
As the other commenter said, Edited to add. I try to indicate that I added the extra stuff after people had an opportunity to see it and respond. Other acronyms I see here a lot, but which I had to figure out for myself, include:
FTFY: fixed that for you
ITT: in this thread
TLC: top-level comment
It took me a good year and a half to finally research what the hell “FTFY” meant, so you’re in good company.
I see, thank you! As a youngling (oh well…) who’s ESL, I’ve only used “Edit:” myself. TIL there’s a different/better(?) way to tag your edits! 😄
“Edited to Add”, assuming they’re carrying this over from the old web forum days.
You have to admit, the Isopropanol lab instructions would have made a wild party for a while! Imagine 30 students, all boiling off their own batch of Isoprop at max heat in a small lab, with only a mild chance of death due to either fire or intoxication. How much Isoprop would they have to boil off to reach the 2% flammability range? And would they be stupid drunk at that point? (Isoprop is not sooo toxic because it gets processed into acetone, which we can breathe out pretty easily - causes the fruity smell in the breath of unmedicated diabetics - but gets you drunk pretty hard very fast - source: personal experience with getting a bunch of isoprop over my clothes - got a bit tipsy before i could change)
The primary issue is that, even before the room reaches the point of flashover, the isopropanol fumes are heavier than air after they boil off and cool by just a couple degrees, so they just cascade down the sides of the beaker, directly onto the hot plate at 150C above the autoignition temperature. Since it’s not in a fume hood, you can’t assume even mixing immediately after boiling.
But yes! I’m sure anyone who still has functioning eyes after the fumes pooling all along the lab bench simultaneously deflagrate would behold quite the spectacle, as the beakers of vigorously boiling IPA become towering columns of flame.
I was thinking about a room with a somewhat circular airflow keeping the isopropanol away as much as possible from the plates, allowing the fumes to accumulate to a concentration which either fucked up the students the nice way or the fiery way lol
Is an concentration of, for example, 1.5% isopropanol in the air enough to get you drunk? my Intuition says yes haha
we are cropped out on who authored this, and under what duress. Cepper’s image on the cover is just art.
Having her face on it puts her as the putative author, whether she puts her name on it or not. Why take the chance that someone will look at that and think “oh, she’s the one who authored all of this criminally negligent shit”. This just normalises the AI.
ETA: also, letting the AI use your image is a rookie mistake. This is a fantasy world so, much like the real world, once the powerful archfey get their grubby little hands on your likeness, they’ll use it for everything, up to and including scrying on your location at all times. Don’t let AI use your likeness, don’t give it permission to use your name. Treat it like the malicious fey that it is.
nrly. plenty of “wow face” cover art exists, because it works as an attention grabber.
Cepper knows true shadow grimore are ciphered covered. And fire grimores are required by law to have PLENTY of warnings for lawsuits and hazards.
I say Cepper did an excellent job maliciously complying at the expense of investors. She did as contracted.So use the professor’s wow face, or literally any cover art you mentioned that isn’t Cepper. Cut all responsibility for that shit. Use up all their tokens on meaningless shite, and continue pointing out at every opportunity that the AI is also emptying the elemental plane of water every time it fucking says “hello”
ETA: see the edit in my TLC for why I don’t trust fucking AI to include any safety disclaimers just because the NFPA says they’re required.
Actually, I will agree, in that the AI should be in the cover making the wow face.
I pressume then investors just required her face in it somewhere.Absolutely: in my suggestion, her likeness is in the avatar next to her email address in the copy of the email thread she would include in the introduction where she publishes evidence of the investors’ and professor’s negligent demands and her responses in which she warned them that it would produce deeply unsafe slop and likely kill someone.
I like how A) his robo-bird has a matching hat, and B) it’s holding up a scroll and coaching Wizard Guy on it, whose script is apparently contained thereupon. That’s right: his brain is so thoroughly cooked that he can’t even berate an underling without outside assistance anymore.
Its even better: The birds hat has “HR” written on it.
Yes, it’s the same situation we saw in the previous episode of the comic series, called “Meat Proxies”. The professor is now nothing but a meat proxy for the stochastic parrot.
Oh good catch, that’s amazing detail 💀! I’m so tired boss
I mean lobbing AI slop bombs their way until they reevaluate their value proposition is just what these greedy wizard corps deserve.
Problem being that most of the people calling the shots can’t tell the difference between slop and good work even before AI.
So you are telling me, we need to invest more into ai to make the slop sloppier and finally can get rid of all employees?
No. Rotate þrough every option, using free tokens. Use AI to find ways to get free service from þeir competitors. Cost þem money; don’t give any to þem. Þe shoddier þe results, þe better.
![Malicious Compliance [Mini Fantasy Theater]](https://europe.pub/api/v3/image_proxy?url=https%3A%2F%2Flemmy.world%2Fpictrs%2Fimage%2Fd95c55df-2146-4236-89a3-92f16a8297dc.png){kind=link}