Zompist Bboard Again

Arzena wrote: ↑Fri Nov 21, 2025 6:02 pm

Raphael wrote: ↑Wed Nov 12, 2025 11:21 am I could have asked about this in the Linguistic Miscellany Thread, but first, there's a fairly intense and long-winded discussion going on there right now, and second, it's a request for help with a small detail of one of my conworlding projects, s I'll put it here.

I want to make a flag in my future Earth scenario a bit less boring than it currently is, and therefore, I'd like to know if there'd be a good way to translate the slogan "Freedom Kindness Wisdom" into 1) Standard Chinese and 2) Standard Arabic.

In Standard Arabic it would be: الحرية والإحسان والحكمة (al-hurriyya w'al-ihsaan w'al-hikma)

Thank you!

Raphael wrote: ↑Sat Nov 22, 2025 6:44 am

Arzena wrote: ↑Fri Nov 21, 2025 6:02 pm In Standard Arabic it would be: الحرية والإحسان والحكمة (al-hurriyya w'al-ihsaan w'al-hikma)

Thank you!

A warning here: if you plan on pasting this into a graphics program, make sure the program can handle Arabic. The symptom is disconnected letters, possibly in backwards order— not infrequently this makes it into production. If it looks like it does on the ZBB, you're fine.

zompist wrote: ↑Sat Nov 22, 2025 3:24 pm
A warning here: if you plan on pasting this into a graphics program, make sure the program can handle Arabic. The symptom is disconnected letters, possibly in backwards order— not infrequently this makes it into production. If it looks like it does on the ZBB, you're fine.

Thank you again! I think it'll be a while before I try to do any graphics on that base.

However, are your Chinese characters right-to-left or left-to-rich, and which of these should they be?

Raphael wrote: ↑Sat Nov 22, 2025 3:29 pm

zompist wrote: ↑Sat Nov 22, 2025 3:24 pm
A warning here: if you plan on pasting this into a graphics program, make sure the program can handle Arabic. The symptom is disconnected letters, possibly in backwards order— not infrequently this makes it into production. If it looks like it does on the ZBB, you're fine.

Thank you again! I think it'll be a while before I try to do any graphics on that base.

However, are your Chinese characters right-to-left or left-to-rich, and which of these should they be?

I believe left-to-right is standard these days, and right-to-left is found only in special contexts. Top-to-bottom is (I think) still common in Taiwan but not on the mainland.

Actually, on further thought… it’s very possible that a non-transparent flag might have writing going away from the flagpole, so left-to-right on one side and right-to-left on the other. I don’t believe Chinese script is used very much on flags, though.

I'm assuming that the flagpole is on the left side. Thank you!

zompist wrote: ↑Sat Nov 22, 2025 3:24 pmA warning here: if you plan on pasting this into a graphics program, make sure the program can handle Arabic. The symptom is disconnected letters, possibly in backwards order— not infrequently this makes it into production.

Seeing an MCU movie, of all things, get that wrong was extremely disappointing. (ISTR it was Black Panther, which is even worse given that they used Xhosa in it.)

Thank you very much again, zompist and Arzena! The result, for now, is here: https://www.verduria.org/viewtopic.php?p=102285#p102285

Out of curiosity, I uploaded some of the Hyperborea materials and asked Gemini 3 to create a generic image of the conworld:



This is non-canonical because some of this stuff consists of LLM hallucinations and it filling in details I didn't describe. However, it does depict some things I haven't mentioned yet like the Šaké, the priest's dagger.

Unfortunately, using AI for image generation is leeching off of real artists that put their blood, sweat and tears into their art (not to mention AI is burning the planet). So please don't use AI if you can at all avoid it.


JAL

jal wrote: ↑Fri Nov 28, 2025 3:10 am Unfortunately, using AI for image generation is leeching off of real artists that put their blood, sweat and tears into their art (not to mention AI is burning the planet). So please don't use AI if you can at all avoid it.


JAL

I don't plan on using commercial LLMs for my conworld. I would also advise others to avoid them. If I had my way, I would ban their use for commercial products.

People are saying Gemini 3 is the best ever (like they always do), so I wanted to try it out. Upper estimates I've seen for generating one image using state of the art AI: Powering a 9W light bulb for 10–20 minutes, charging an average phone to 15%, streaming 1–2 minutes of an HD movie on a TV, running a 1000W microwave on high for 5–10 seconds, driving 25–50 feet in an EV (maybe driving an extremely efficient EV 50–100 feet in an ideal environment) or 2.5 human calories (about 1.5 grapes).

In other words, the energy consumption is higher than end users might imagine, but generating one image is nowhere near world-endingly bad. You should talk to the marketers who generate images around the clock (~129,600 grapes a day). The absolute prohibition on using AI is the left's new mass hysteria.

rotting bones wrote: ↑Fri Nov 28, 2025 3:37 amThe absolute prohibition on using AI is the left's new mass hysteria.

Ah, yes, blame "the left". I realize this isn't a political site or, god forbid, X, but blaming "the left" for something is just absolute brain-rotting (though not quite bone-rotting) dumb.

Also, you focussed only on the energy consumption, something I put between parenthesis, and totally bypassed the fact that you're indirectly using stolen artwork from real artists. That's bad dude, just bad.


JAL

jal wrote: ↑Fri Nov 28, 2025 3:59 am Ah, yes, blame "the left". I realize this isn't a political site or, god forbid, X, but blaming "the left" for something is just absolute brain-rotting (though not quite bone-rotting) dumb.

It is the left. The right loves AI and wants to use it to destroy the planet.

Responding to actual data by saying it's "dumb". Not symptomatic of mass hysteria at all.

jal wrote: ↑Fri Nov 28, 2025 3:59 am Also, you focussed only on the energy consumption, something I put between parenthesis, and totally bypassed the fact that you're indirectly using stolen artwork from real artists. That's bad dude, just bad.

Gemini's Nano Banana is not trained on images indiscriminately scraped from the web. The images it is trained on are either public domain or licensed for use in AI training. In fact, Google will defend you legally if you are sued for using one of Gemini's images.

I didn't address this because it doesn't matter. Like I said, I'm not using the image commercially and I urge others not to.

rotting bones wrote: ↑Fri Nov 28, 2025 3:37 am The absolute prohibition on using AI is the left's new mass hysteria.

"hysteria" is a word that people usually use when a lot of other people have a particular opinion or idea or stance, and they themselves really don't like that opinion or idea or stance, but they can't think of any actual good arguments against the opinion or idea or stance. Then, they call it "hysteria" and act as if, by calling it that, they would somehow have magically refuted it.

See also: "absurd", although in that case there's less of a requirement that the opinion or idea or stance must be shared by a lot of people.

Raphael wrote: ↑Fri Nov 28, 2025 4:28 am "hysteria" is a word that people usually use when a lot of other people have a particular opinion or idea or stance, and they themselves really don't like that opinion or idea or stance, but they can't think of any actual good arguments against the opinion or idea or stance. Then, they call it "hysteria" and act as if, by calling it that, they would somehow have magically refuted it.

See also: "absurd", although in that case there's less of a requirement that the opinion or idea or stance must be shared by a lot of people.

I have refuted it here using facts and logic.

I don't know if AI makes sense in conworlding. Isn't creating things in the most painstaking manner imaginable part of the appeal?

rotting bones wrote: ↑Fri Nov 28, 2025 3:37 am In other words, the energy consumption is higher than end users might imagine, but generating one image is nowhere near world-endingly bad. You should talk to the marketers who generate images around the clock (~129,600 grapes a day). The absolute prohibition on using AI is the left's new mass hysteria.

The tactics are debatable but the sentiment is correct.

The per-image cost is small, but aggregate demand must be staggeringly high because Google (or is it Microsoft? I think it's Google) is considering nuclear reactors in datacenters.
In an ideal world, the proper response to that should be 'what the hell are you thinking?' or even better the question would never arise because tech companies shouldn't have anywhere near the kind of money it takes to put nuclear reactors in datacenters. Yet they do, and governments think it's swell, so here we are.
It always comes down to blaming the end user, which is wrong, but it comes from a place of desperation, not hysteria. End users are the only people who could possibly listen.

Ares Land wrote: ↑Fri Nov 28, 2025 4:34 am The tactics are debatable but the sentiment is correct.

What is correct, that I shouldn't generate one image to try out a commercial LLM?

rotting bones wrote: ↑Fri Nov 28, 2025 4:38 am

Ares Land wrote: ↑Fri Nov 28, 2025 4:34 am The tactics are debatable but the sentiment is correct.

What is correct, that I shouldn't generate one image to try out a commercial LLM?

I'm speaking in very general terms -- the current fad for LLMs is a problem. It pains me to say it, because I'm a technophile at heart,but it is.

I don't think trying out LLMs from time to time is wrong, in fact I do it myself.

Ares Land wrote: ↑Fri Nov 28, 2025 4:43 am I'm speaking in very general terms -- the current fad for LLMs is a problem. It pains me to say it, because I'm a technophile at heart,but it is.

I have already urged people not to use commercial LLMs professionally. I hate the AI industry. I sincerely hope their bubble pops this time.

Ares Land wrote: ↑Fri Nov 28, 2025 4:43 am I don't think trying out LLMs from time to time is wrong, in fact I do it myself.

I have used open source SLMs running on a 16 GB graphics card to find inconsistencies in long documents, including for Hyperborea. Without SLMs, it consists of years of inconsistent, unedited notes. I can't always find the inconsistencies myself, and the texts are too long for me to expect someone else to edit it for free.

jal wrote: ↑Fri Nov 28, 2025 3:59 am Ah, yes, blame "the left". I realize this isn't a political site or, god forbid, X, but blaming "the left" for something is just absolute brain-rotting (though not quite bone-rotting) dumb.

I apologize if I was rude. I have fallen for mass hysteria myself. I didn't intend to attack you. From my perspective, I was warning you not to fall for popular exaggerations.