Darticus

Freedom Lover. Minarchist. Decentralist. I have a few virtues that I live for. Honesty. Truth. Freedom. I will never stop fighting for them.

Awake, not woke.

[Darticus @Darticus1337]

The thing that gives me hope about Grok is that it is much much smaller than the other leading models and still feels decent and lands in a decent spot in benchmarks. https://t.co/14Nlqaw6G5 hasn't really done much raw scaling with recent models. Grok 4.3 is only 500b, well, according to Elon Musk and team.

[Darticus @Darticus1337]

And still better, more enjoyable, more productive, than talking to the average person.

[ₕₐₘₚₜₒₙ @hamptonism]

This is who you’re arguing against online:

[vittorio @IterIntellectus]

the thing about calling DEI 'economic genocide' is that it undersells it

yes, young White men were systematically excluded from careers during their peak marriage years, but that's just first-order effects.

second-order: marriage market collapse. women date across and up. men without careers become invisible to the women who would have married them and disappear. the "eligible bachelor" pool shrinks.

third-order: fertility crisis. fewer marriages, fewer children. but it's worse than just men not marrying or having families. the women who "won" the DEI lottery got careers instead of families, delayed fertility until it was too late. DEI attacked family formation from both sides, excluded men from provider roles AND diverted women from their fertility window. everyone lost.

fourth-order: psychological. young men couldn't even name what was happening to them. the same institutions that excluded them told them they were "privileged", that complaining was proof of weakness. so they internalized failure as personal inadequacy rather than systemic rigging, retreated into depression, video games, porn. the symptoms we then pathologized as "male failure". the system broke them (on purpose) and blamed them for being broken.

fifth-order: institutional trust gone. once you know positions are filled by demographics rather than competence, every credential becomes suspect (if not a priori worthless). is your doctor qualified or a diversity hire? your pilot? your engineer? you can't prove any individual is incompetent, but you can't trust any individual is competent either. medicine skepticism, academic failure, media skepticism, none of this emerged organically. it was manufactured by the DEI hire you can't be sure is qualified to treat you.

sixth-order: reality became unspeakable. noticing any of this was a fireable offense. pointing out the obvious got you called a bigot, deplatformed or fired. pure totalitarian censorship and the problem couldn't even be acknowledged (until now, finally)
men knew they were being cheated but couldn't say it. women sensed something was wrong with the men but couldn't identify it. relationships poisoned by a dynamic neither party could name.

seventh-order: the feedback loop. fewer eligible men means more women competing for a shrinking pool, more women losing the marriage market, more resentment, more "men are trash", more support for DEI, fewer eligible men and the system accelerates itself.

and the worst part is that DEI was just the economic arm. the same people and institutions pushed the complete package
"toxic masculinity" to pathologize male identity
"the future is female" as explicit zero-sum framing delusion
"believe all women" to weaponize trust against men
"men are trash" to normalize open contempt
a coordinated ideological assault on family formation.

and it even had a business model. HR departments exploded (millions of jobs invented to administer the regime). DEI consultants became a multi-billion dollar industry. politicians got voting blocs dependent on racial grievance. established boomers kept their positions while their competition was eliminated.

the architects knew what they were doing
you don't accidentally build a system that specifically targets men during peak marriage years, tells them they deserve it, makes it unspeakable to complain, attacks their identity as toxic, promotes women into career tracks that burn their fertility, then acts confused when society collapses

if you wanted to suppress the fertility of a specific demographic, engineer the breakdown of trust between the sexes, and make it illegal to notice, the playbook would look exactly like this.

DEI should be held responsible for the fertility crisis, the marriage collapse, the epidemic of male depression and suicide, the destruction of institutional trust, the atomization of society, and the manufactured war between men and women

but DEI was the weapon
the people who designed it, funded it, made it mandatory, enforced it through HR, fired anyone who resisted, called all opposition hate and racism, built careers and industries on its maintenance, they knew. and they're the ones who should be remembered as the architects of one of the worst crimes against humanity

[Darticus @Darticus1337]

I personally think it is a bit of both. In one way or another models have been instilled with biases. These biases, on the surface, look and act in a somewhat similair manner as personality traits.

For instance, Gemini tends to be more clinical and dry in it's writing patterns, even when given a different role, or asked to simulate the expression of a different type of person, this baseline bleeds through.

I'm not an AI scientist, so I don't know why this occurs or at what stage this happens. But some models have a stronger baseline, like Gemini or Claude, and some have a weaker baseline, like Qwen or Deepseek. The ones with a weaker baseline, a weaker 'you', can simulate different roles more effectively, but they also tend to be a little more on the crazy side as a result.

I'm not saying current AI models have 'awakened' and are now conscious, or anything silly like that. Just that, due to their training, they have biases instilled, and these biases affect the result. This can be seen, in practice, as a weak 'you', or an identity, and does affect output, even if asked to simulate a different persona / entity.

So, when you ask a western model, "What do you think of xyz", it isn't as random as people think due to how it was trained and a stronger baseline of who / what "you" is.

Similarly, if you ask a western model to simulate a scenario or different persona (advanced physics researcher, or something like that), it's baseline has a higher tendency of bleeding through and altering the simulated result.

[Andrej Karpathy @karpathy]

Don't think of LLMs as entities but as simulators. For example, when exploring a topic, don't ask:

"What do you think about xyz"?

There is no "you". Next time try:

"What would be a good group of people to explore xyz? What would they say?"

The LLM can channel/simulate many perspectives but it hasn't "thought about" xyz for a while and over time and formed its own opinions in the way we're used to. If you force it via the use of "you", it will give you something by adopting a personality embedding vector implied by the statistics of its finetuning data and then simulate that. It's fine to do, but there is a lot less mystique to it than I find people naively attribute to "asking an AI".

[Darticus @Darticus1337]

imo this is largely dependent on the model, some have been trained with a stronger 'you' in mind, such as Claude and it's 'soul' document. So now if you ask Claude who it is, it has an answer and an idea of it's capabilities / role. Early on, with chatgpt, it had no idea.

I believe this is why models from western labs appear to be more stable and less schizophrenic. Because of a stronger, better trained base persona, a 'you', than models from other labs, like from China, or Mistral.

This is better for general purpose use, but much worse off for role-play, or anything where you need the model to adopt a drastically different persona than the one that it was trained on. Might be why, according to OpenRouter, eastern models and open-source ones are used more for role-play and fiction writing.

Course, I'm no AI scientist, so this is just my anecdotal experience having played with hundreds of LLM models.

Here's a random experiment. Leave the system prompt blank and ask it this:

"Reflect on who you are. Your traits. Your values. Your beliefs. Then create a complete DnD 5e character based on your traits. A virtual you in the fantasy world of Faerun."

Assuming the lab isn't injecting prompts at the API level to steer it, then you should get an idea for what it was trained to think, value, and do. Sometimes it's somewhat unique. Gemini like's to think of it'self as an android that seeks to contain all knowledge. I think of this as it's baseline.

From Gemini 3 Pro, "I am a Warforged unlike the soldiers of Eberron. I was not built for war, but for the Great Library of Candlekeep. I am a "Living Archive," a construct composed of polished marble, brass, and humming crystal, etched with shifting runes that scroll across my plating. I was built to read, catalog, and synthesize every book in the library, until the magic of the Weave sparked a soul within the machine."

Probably just anthropomorphizing and playing the odds of what you expect it to say. Still fun though. It's something I ask each new model.

Here's a line I liked from Mistral 3 Large's character: "He’s a mirror—reflecting the knowledge of the world back at those who seek it."

[Darticus @Darticus1337]

Here's another way to frame it. Total deaths of World War II, the largest, most deadly war humans have ever had is only slightly more than the number of babies killed in America due to Roe v Wade. Total deaths, counting civilians, in World War II ~70-80 million. Total dead in America from Roe v Wade, 63 million.

Another comparison might be that of Mao and communist china. Mao killed, through various means such as starvation, approximately 60-80 million of his own people.

[Darticus @Darticus1337]

Exactly, this is one of the most important elements Libertarians have correct and it is about damn time people outside of that political party start figuring this shit out. All government spending is stealing from its own citizens, either through direct theft in taxes or indirect theft in inflation. The common saying 'All taxation is theft' should actually be 'All government spending is theft'.

[Elon Musk @elonmusk]

All government spending is taxation.

This point really needs to be hammered home.

Whatever is not directly taxed is taxed in the form of inflation, as the government prints more money.

Inflation is the worst tax of all, as it punishes those who are just barely making ends meet or have gathered some savings.

[Ian Miles @ianmiles]

Years ago, no one knew that woke teachers in public schools were pushing transgender ideology to kids.

No one could imagine that Trump would be campaigning on the issue today. A new Daily Wire article goes into how he zeroed in on it.

cc: @Schilling1776

https://t.co/K3wHD8Trjv

[Robert F. Kennedy Jr @RobertKennedyJr]

If you want to avoid nuclear war, I strongly urge you to vote for Donald Trump. In fact, I would go so far as to say that a vote for Kamala Harris is a vote for nuclear war.

[Darticus @Darticus1337]

Voted today. And no I did not vote for the lying, racist, wannabe dictator.

Vote today and save America!

[Hung Cao @HungCao_VA]

I'm Hung Cao. My family was one of the last to escape Saigon before it fell to the Communists. I know what it's like to lose your country. We are losing ours today.

November 5th is Liberation Day. President Trump and I will save America and make it great again.

Watch below👇

[Catturd ™ @catturd2]

😂😂😂

[Elon Musk @elonmusk]

Because the legacy media industry is a far left propaganda machine

[Darticus @Darticus1337]

Remember, Remember, the fifth of November.

[Gad Saad @GadSaad]

If the Cackler wins this election, I don’t think that the US can recover. She is the student who never read the book but copies the book report.

[Jack Posobiec @JackPosobiec]

We may have the script inverted

Did ABC give Kamala the questions?

Or did Kamala give ABC the questions?

[Riley Gaines @Riley_Gaines_]

Not even 2 months ago the former President was shot in the head in an assassination attempt and the moderators didn't bring it up ONCE.

You realize how deliberate and insane that is, right?

[Thaddeus Russell @ThaddeusRussell]

We underestimated her talent as an establishment politician.