lvxferre

As another user and me were discussing, in another thread about the same topic, I believe that the 10% admix is likely due to coastal settlements here in South America. Nothing too fancy, just to facilitate trade. Specially with the folks in Central Andes, as the Andeans had a good and large (albeit land-based) trade network already. And, well, when you run this sort of settlement you're bound to interact with locals, right?

So this is just another part in reducing cost on section that doesn’t produce money.

That's what I immediately thought - they're cutting corners to decrease dependency of googlebux, as depending on how things go those bux will go dry.

But there’s enough of a problem you can see even if you just start at Julius, which is what I was concentrating on in my previous comment. The parallels to Trump are terrifyingly on the nose.

True that.

Weirdly enough (or perhaps not surprisingly) I see the same here with Bolsonaro supporters; there's a disproportionally high amount of them among classicists, even if humanities as a whole leans heavily to the left.

The alt right obsesses over the Roman empire, but ignores the republic, as if Julius Caesar and Octavius were the origin of everything. As such I'm not surprised that they don't learn about what caused the fall of the republic. (A century or so of oppressed masses and greedy elites did it.)

And, even when it comes to the empire, they're busier cherry-picking examples that show that the grass was greener, the men were manlier, the women were chaster, and dogs barked quieter.

At least when it comes to languages, the eurocentrism and subjectivity are being addressed for at least a century. Sapir for example proposed that the "classical languages" weren't just two but five - Arabic, Chinese, Greek, Latin, Sanskrit. And the definition became roughly "varieties with a heavy and outlasting impact outside their native communities". (Personally I'd also add Sumerian, Quechua and Nahuatl to that list. But that's just me.)

Additionally plenty linguists see the idea of "classic" not as specific languages, but as a potential stage of a language, assigned retroactively to the period when its prestige and cultural production were specially strong. For example, Classical Ge'ez is defined as the one from centuries XIII~XIV.

They probably could, indeed - but you'd need multiple different applications, each for one use case. In the meantime a LLM offers you a tool that won't hit all the nails, or screw all the screws, but does both decently enough in the lack of both a hammer and a screwdriver.

It's a great analogy though - Linux users aren't deemed profitable by the A³ companies, just like offal is unjustly* deemed yucky by your typical person.

*I do love offal though. And writing this comment made me crave for chicken livers with garlic and rosemary over sourdough bread. Damn.

The backlash to this is going to be fun.

In some cases it's already happening - since the bubble forces AI-invested corporations to shove it down everywhere. Cue to Microsoft Recall, and the outrage against it.

It has virtually no non-fraud real world applications that don’t reflect the underlying uselessness of the activity it can do.

It is not completely useless but it's oversold as fuck. Like selling you a bicycle with the claim that you can go to the Moon with it, plus a "trust me = be gullible, eventually bikes will reach Mars!" A bike is still useful, even if they're building a scam around it.

Here's three practical examples:

1. I use ChatGPT as a translation aid. Mostly to list potential translations for a specific word, or as conjugation/declension table. Also as a second layer of spell-proofing. I can't use it to translate full texts without it shitting its own virtual pants - it inserts extraneous info, repeats sentences, removes key details from the text, butcher the tone, etc.
2. I was looking for papers concerning a very specific topic, and got a huge pile (~150) of them. Too much text to read on my own. So I used the titles to pre-select a few of them into a "must check" pile, then asked Gemini to provide me three paragraphs summaries for the rest. A few of them were useful; without Gemini I'd probably have missed them.
3. [Note: reported use.] I've seen programmers claiming that they do something similar to #1, with code instead. Basically asking Copilot how a function works, or to write extremely simple code (if you ask it to generate complex code it starts lying/assuming/making up non-existent libraries).

None of those activities is underlyingly useless; but they have some common grounds - they don't require you to trust the output of the bot at all. It's either things that you wouldn't use otherwise (#2) or things that you can reliably say "yup, that's bullshit" (#1, #3).

I would like to introduce you to the indie game scene. Where AAA is faltering, indie has never been in a better place.

Amen.

Indie games might not be flashy, but they're often made with love and concern about giving you a fun experience. They also lack all those abusive DRM and intrusive anti-cheat systems that A³ games often have.

It's interesting how interconnected those points are.

Generative A"I" drives GPU prices up. NVidia now cares more about it than about graphics. AMD feels no pressure to improve GPUs.

Stagnant hardware means that game studios, who used to rely on "our game currently runs like shit but future hardware will handle it" and similar assumptions get wrecked. And gen A"I" hits them directly due to FOMO + corporates buying trends without understanding how the underlying tech works, so wasting talent by firing people under the hopes that A"I" can replace it.

Large game companies are also suffering due to their investment on the mobile market. A good example of is Ishihara; sure, Nintendo simply ignored his views on phones replacing consoles, but how many game company CEOs thought the same and rolled with it?

I'm predicting that everything will go down once it becomes common knowledge that LLMs and diffusion models are 20% actual usage, 80% bubble.

I don't see this as an unpopular opinion, but I do agree with it - at least here (Brazil) Twitter was evolving into a containment cage for nutjobs and morons, until it was blocked. (And it's damn easy to find who's who in the Bluesky diaspora, as the nutjobs/morons miss Twitter while the saner people are glad to see it locally gone.)

This is the sort of thing that I love reading on the internet.

From a conlanger perspective I feel like the time reference could be split into four, to account time travel. For example: let's say that both of us travelled to 3100, I remained there and you came back to 2024. Then you write me a letter, that I'm going to read as soon as we arrive in 3100, telling me about your experiences. You could use:

• your current date as reference - 3100 comes after 2024, so it's future
• your personal experiences - you already experienced it, so it's past
• my current date as reference - as I'm in 3100, it's present
• my personal experiences - as I'm watching you experience it, it's present

Any given language could pick any of those references to model their tense around, or many of them, or even none (plenty languages IRL lack grammatical tense). If only doing things from the PoV of the speaker (you), that means 6~9 tenses for what most languages have 2 (past and non-past) or 3 (past, present, future).