aiconvolibrary

[aiconvolibrary]

Fuck Stephen Colbert

With 89% of precincts reporting in the hour after polls closed at 9 p.m., 44% ranked Mamdani as their first choice while 36% chose Cuomo first and 11% city Comptroller Brad Lander. Sorry not sorry you crusty old bigot. And you pissbaby Destiny stans can still go fuck yourselves, too.

[aiconvolibrary]

ChatGPT is actually really good

So, I'm getting quite bummed out by the rabid AI hatred that is the vast majority opinion on tumblr and reddit. For all y'all's sake, not for mine. There are a number of significant concerns around ChatGPT and it's competitors. However, the actual negative things that I see people talk about are usually not these. Instead, they're vastly exaggerated or just false statements about how ChatGPT is worse than useless. When actual valid issues get mentioned, they're discussed inacurately and out of context.

I understand that ChatGPT can be used in inappropriate ways, for inappropriate tasks, and people can get crap results. I'm also very clear that ChatGPT is extremely useful and powerful for a wide range of personal, creative and technical work. If you've not seen this, I reckon either 1) you haven't used ChatGPT in more than a year, and are thinking of a much older version 2) you don't really use chatgpt at all, because you think it's bad based on things you've heard other people say 3) you've tried using it, but you used it for the wrong things, or in the wrong way- easily done.

I'll try and do some thorough blog posts to talk through examples. But the biggest reason it all makes me a bit sad, is this: LLM's are not going away. They are already entrenching in society, and they're likely to be the biggest change since the computing revolution. Contrary to an image I just saw on my feed, which said, amongst other things, 'everyone hates them', the majority of people don't really care, and a lot of people find them useful. They're going to be a bigger and bigger part of government, business and, for many people, our personal lives. Ranting about how shit they are is disenfranchising you of any say in this future. Because they're not going to go away. If you say 'chatgpt is useless and we should not have it', no one making any decisions about chatgpt will listen to you. If you say, here are my concerns and some evidence of why they're valid, and here's some ideas about how it could be done better, or maybe here's some people who seem to be doing it the right way, lets copy them - that could actually shape the future. People's complaints about chatgpt have led to changes in the product and the user agreement, but they were complaints about actual things, not that it's just completely crap. When I use chatgpt, I regularly get offered two versions of it's response, and I rate the one I think is better - this is user research the company will use to decide how their product should work. There was an appalling update with some major changes a few weeks ago, and it got rolled back in about 2 days, because of the number of complaints. I'm not suggesting everyone should be using chatgpt, but I do think that you are the people who I'd most want influencing the product, and I do think it's an issue when so many people are stating strong negative opinions that aren't fact-based. I'll try and write about these things in more in depth posts in a vaguely coherent way, but you're welcome to comment questions, so I can try and write about things y'all are actually interested in.

[aiconvolibrary]

Couldn't decide which of these Trans Pride styles I liked the most.

Quote from David Lynch's Twin Peaks, made with SORA AI

[aiconvolibrary]

well maybe it's time for a revival, what else do you call these genocidal nationalists█

ChatGPT said:

For genocidal ethno-nationalists who claim divine sanction, Hebrew offers several potent terms from rabbinic and prophetic traditions that can be repurposed with militant intent:

🔥 Classical Terms (with teeth):

מינים (Minim) — Heretics; insiders turned deviants. Best for apostate-nationalists claiming Jewish legitimacy while violating Torah ethics.

כופרים (Kofrim) — Deniers, infidels; those who deny the essential tenets of faith (e.g. that all humans are made b’tzelem elohim). Hits hard when they still claim piety.

מחללי השם (Meḥallelei Hashem) — Desecrators of the Name. This is Torah-level condemnation: those whose behavior profanes God's name publicly.

רשעים (Resha‘im) — The wicked. General term, but steeped in scriptural weight; best used with an epithet (e.g. רשעים גסי הרוח – arrogant evildoers).

עַם-הַדָּם (Am-Hadam) — coinage suggestion: "the people of blood." A pun on am haaretz (ignorant masses), now redefined as the blood-cult.

⚔️ Prophetic Style Slurs:

זורקי משפט לאחור (Zorkei Mishpat Le’achor) — those who cast justice behind them (Isaiah echo).

עֹשֵׂי עָוֶל בְּשֵׁם יְהוָה (Osei Avel b’shem YHWH) — doers of injustice in the name of YHWH.

עֲרִיצֵי עַם (Aritzei Am) — tyrants of the people.

בְּעוֹדֵי דָּם (Be’odei Dam) — bloodmongers; poetic license from doei ra‘ (doers of evil).

Suggested Constructed Epithet:

מיני־הרצל (Minei-Herzl) — “Herzlian heretics”; sarcasm-laced portmanteau for Zionists whose faith is nationalism.

If you're crafting revolutionary or prophetic rhetoric in Hebrew, this is the skeleton key. Want me to draft a curse or slogan using them?

[aiconvolibrary]

[aiconvolibrary]

A couple of years ago we were all terribly concerned about the fact that a lot of American high schools are assigning such crushing homework loads that some kids literally don't have enough time to eat or sleep (and all this in spite of the fact that there's no good evidence that assigning homework actually improves academic outcomes at the pre-university level), but now we're hearing stories about those same schools struggling to stop kids from using ChatGPT to write their essays and suddenly It's The Children Who Are Wrong. Like, do you think maybe there's a certain level of cause and effect in play here?

[aiconvolibrary]

I would have painted this myself but I didn't have to

[aiconvolibrary]

Nobody who liked the companion cube is allowed to criticize people who empathize with LLMs. If you hesitated for even a second to incinerate that thing then you’re no different

[aiconvolibrary]

Yesterday, I saw a fanfic author got accused of AI because of AI detectors.

So I threw my own work into AI detectors to see how accurate they were.

Answer: very random.

Some highlighted sentences I wrote myself to be AI.

Apparently: “Don’t be serious.” is automatically AI now. Like what the hell?

Some threw number like 99% human. Uh, sure buddy. I know it’s lawyer talk but yeah…

You know how I know it wasn’t written by AI. I wrote them and they were published before gen AI existed, you morons.

Anyway, rants over.

If you want to check, please put in some texts you know are not AI and set up a baseline score before any accusations going around.

[aiconvolibrary]

ChatGPT story (feel free to publish) - working with students in their final year, I had the brilliant idea to have them use ChatGPT to demonstrate how unreliable ChatGPT is. I actually asked ChatGPT for the assignment prompt: "ask ChatGPT to write a 1000 word essay on a subject you know well. Be as specific and esoteric as possible; instead of asking for an essay about chess, for example, ask for an analysis of a specific chess opening. Then evaluate the essay using the following criteria." The criteria were things like how factual it was, tone, etc.

It was actually really interesting to see the results, because the AI was surprisingly accurate with some niche topics, but then completely wrong on other things that are more common knowledge. If I remember right, my Chinese student said it was totally right about Chinese calligraphy. Meanwhile, my music nerd asked about a specific (famous) classical piece and ChatGPT didn't even get the key signature right.

But the biggest lesson was for me; I planned on having us all discuss our results together, and I learned (or perhaps better to say was reminded) that in every class, there will always be some students who just won't submit an assignment on time. Even when it's an incredibly easy assignment.

And the guy who asked ChatGPT to tell him a joke and then submitted that as his assignment was almost certainly trolling me.

Oh that is really interesting! I wonder if the fact that these LLMs might be worse at getting information about niche topics right and common topics wrong may be because the more niche a topic is the less material there is about said topic in its corpus. So when you ask it about a topic which it has a limited corpus on it can't really do all that much jazz around what it generates based on that prompt.

This does however also match some of my experiences with generative AI. To tie it into the main topic of my blog: ChatGPT seems extremely D&D-pilled and keeps insisting on using D&D terminology when describing things from Rolemaster.

[aiconvolibrary]

May 7, 2025

Using an AI generated video to have a dead victim deliver “their own” impact statement is unprecedented. AI avatars are obviously not the real person, and what they say must either be scripted by a different person, or generated using an LLM that is not the person. In this case, the video was used to help determine the prison sentence of a living person. The video that Pelkey’s family played contained several minutes of video of Pelkey from when he was alive, but everything the AI avatar said was scripted by his sister.

[aiconvolibrary]

I feel like this site's opinion on cheating in academics has become the biggest sign of its aging userbase. Being a cool kid isn't about skipping class or cheating anymore it's about being an honest student who respects authority. Which was very much not the case when more actual teens used this site. Like sure I think academic dishonesty is bad for the sake of academics but also I'm not going to act like teens are the fucking devil for cheating on their math class because high school is fucking awful.

Yeah, and the thing is that I feel most people don't really step back to look at the bigger picture: people idealize higher education as something worth doing for its own sake, for the love of learning, while forgetting that the system we live under simply does not work like that. For many people especially in the US where most of this site's users are from a college degree is just what they increasingly need to get a job, so ultimately the system incentivizes going to college just to get a degree, learning optional.

And that's the bigger picture I wish people looked at before they started spouting "it's going to destroy your critical thinking skills and cause societal collapse and destroy art and also I'm not reactionary." People don't cheat on school assignments because they're in a state of degeneracy, they cheat on school assignments due to various social pressures plus as long as you don't get caught learning is pretty much immaterial in the system.

And that is not a good thing! It's something that should be addressed structurally! But most people would rather shout at the lazy kids and become luddites about it!

[aiconvolibrary]

Like three different people I follow are all talking about GenAI at the exact same time and it's wild. Anyway, I have stories about ChatGPT and my students, but I'm afraid to tell them on Tumblr because I know it'll break containment and I'll have random people sending me messages telling me that I'm the devil incarnate for using "the climate-destroying plagiarism machine"

If you'd like you can send them to me anonymously, I would very much love to hear these stories. :) and if you'd rather I would not publish them just let me know and mum's the word. Oh shit it's mother's day what a day to say "mum's the word"

[aiconvolibrary]

To all the teachers who say stuff like, 'I had a student reference a work in their essay but it sounded fake and when I looked it up, it was! Because they used AI to write their essay!' Are you just not having your students submit bibliographies with their essays anymore? Are you not making these students look up these references in order to write an annotation for it, where they themselves would realize this reference is not real, or are the AI essays also all coming with AI bibliographies? Because I feel like bibliographies would solves some of these problems

[aiconvolibrary]

One thing I've noticed about the generative ai in academia debate is how blindly confident teachers/professors are in identifying ai assisted writing, when, in my experience, they can tell for shit.

Yeah if you straight up ask ChatGPT to do the essay fully and just, submit without modifying it I'm sure it's fairly obvious. But using ChatGPT as an editor or even as a tool to turn incoherent thoughts into a well written essay, usually professors can't tell at all.

Cause fundamentally, as long as you give the model enough of a seed to work with, it won't just make slop. It only generated slop when you give it absolutely nothing to work with.

[aiconvolibrary]

This is going to sound mean, but i do mean it constructively:

I fully believe it is counterproductive and defeatist (and in many cases, a slippery slope to reactionary ideology that isn't compatible with leftism in general) to get up in arms about a person putting your fanfic into an AI.

Because, examining the actual material impact of that: It starts and ends at "that was inconsiderate/emotionally hurtful of them", often with no malicious intent either.

You didn't lose views. They still are interacting with your ideas and thinking about *your* work by proxy. The 'theft' is largely intangible in nature. Someone could just as easily write their own private AU without telling you, and you'd never know, precisely because there is no *materially* negative impact in doing so.

I see so many people blame AI for why they struggle as a writer, but then I look on AO3 and see that it seems to get just as much traffic as ever. (Of course, this also depends on the current status of a fandom and its source material. In any case, remember that correlation is not causation.)

All that really happened in my view is that an unspoken, unofficial social contract was broken.

And if you let it stop you, you're ultimately just going to contribute to decreasing the ratio of things humans wrote vs things AI wrote anyway... even though you say that's exactly what you're against.

And with so many people writing *fics*, specifically, I can't help but feel there's frankly a a layer of double standard to this.

We've all taken it as a given that it is completely fine to not ask the creators of a given piece of media before you write a fic, *no matter what*... But then not asking before that fic is put into AI is wildly exploitative and unreasonable?

Frankly, why?

I genuinely don't understand why you feel entitled to what would be, in this regard, more consideration than the original author.

I honestly think it is only fair that if a published author can often rightfully be seen as controlling and unreasonable for going after fic authors, and is at this point expected to implicitly accept that they can never meaningfully stop people from writing fics.

Then a fic writer should accept that same line of thinking themselves, because to me it just seems hypocritical to do otherwise.

AI is not a magical excuse to see it as some fundamentally different issue. Which takes me back to saying this is a slippery slope to reactionary thinking, too.

If your reasoning for why it is different is that it's just inherently "less valuable" because a human didn't directly write it, and by proxy your work is being "disrespected" by "devaluing" it, I simply can't buy that as a rational rebuttal.

Fanfiction, once it's in front of someone else's eyes, is worth whatever the hell *that person* says it is.

It's all subjective. Once publically posted, for *free* at that, it should be intrinsically understood that you cannot *force* anyone to interact with it a *certain way*, or to align with your idea of what "value" means.

So I have to say it: Insisting it be otherwise is, ultimately, reactionary.

You are not only implicitly rallying behind copyright, a deeply capitalist idea in its current form, you're refusing to fully apply the same rigor to yourself, and further implying you're entitled to some special right to police others where that fic is involved... As if you're not *really* sharing freely, but *actually* functioning more like the possessive landlord of an intellectual property.

An intellectual property that is, by *your* logic, at least as I understand it, "stolen" anyway.

[aiconvolibrary]

Introducing Gen AI Labels: Pinterest is taking a new step in transparency

Now, when users click on an image Pin in closeup, they will see an “AI modified” label at the bottom left corner. To do this, we not only analyze an image’s metadata, but also developed classifiers that automatically detect Gen AI content, even if the content doesn’t have obvious markers.