Honestly? Yeah.

Like, we consider red flags for something being "ai" to be... punctuation? As an old who actually learned how to fucking write a sentence, I've actually taken to just leaving in a lot of the typos from typing too fast as a result. Which sucks because I am actively writing worse to still look "human".

Although the good news is that we are only a few cycles away from AI slop being the kind of shit that makes you wonder if anyone else smell's toast. At which point I can go back to writing "normally".

Still fucking hate that every few weeks there is a new word-a-day calendar word that means "you are AI". But that at least is no different than "ha ha, you type too much" or "you try too hard" as being just the normal strive for mediocrity that has plagued the world for decades.

... Twitter served a similar function, a decade ago, by literally condensing politics into slogans that could fit into a tweet.

Yeah, LLMs are another level of bad, but... we've collectively been at this for a while, stupifying our level and breadth of discourse.

Maybe it's where I browse but I haven't really seen any accusations that well written content is LLM generated. Language models have a rather specific tone they like to use, and it's relatively easy to distinguish once you've learned those patterns.

I have seen, however, blatantly LLM generated text going undetected.

Reject internet

Return to Ham Radios

Also, in the age of online communications, when information can be limitlessly copied, it’s impossible for authoritarians to destroy information they disapprove of. It is possible, however, for them to make it difficult to find by flooding the world with vast quantities of plausible-looking alternatives to it, making it hard to identify what is real. Before LLMs, doing this at the sort of scale that would have been effective was prohibitively laborious.

I don’t think the quality that gives prose an AI smell is eloquence per se, so much as the use of common tropes and rhetorical formulae. The kind of rhetoric AI is best at is the kind that relies on appeal to a shared tradition of discourse more than on novel arguments or reason—and I think that type of rhetoric is more common on the right than the left.

The only people I talk to don’t know shit about AI and think all the dogs saving babies videos are real. I think I’m good.

But yes your point stands and this second guessing is the same kind of shit that stops genuine human connection similar to when people stop and ask who they voted for or see the colour of their skin.

I downloaded 17 years worth of my comments before overwriting and deleting my old reddit account. Been thinking about QLoRA fine-tuning Qwen on those comments. Not for use on the internet or anything, just so I can streamline the process of arguing with myself.

Train an LLM that accuses everyone of Fascism. Should be easy

Uninhibited by erudition or imagination the fascist mind misapprehends the appearance of their own imperfection and woe as the machinations of their propagandized scapegoats instead of their own shortcomings, and are welcome to take a close observation of the umber hue of my lowest sphincter.

Look, this isn't doing them a favor at all. Cultists have always ignored things they didn't like and embraced things that they did. Generative IAI doesn't change that in any way.

Eloquent my ass. There's nothing eloquent about the text output of an LLM. Verbose maybe, overly verbose.

Well in 1934 we had book burnings. Have you considered an LLM burning?

LLMs are doing fascists SO gloriously dirty.

There are hundreds of open source models with historical contexts up to like October. Most are about 3-6 months behind current. They're scraping the whole internet, more or less. That's bad, yeah. Except...

Fascists cannot rewrite history anymore

These hundreds of open source models have downloads in the millions.

The number of open source LLMs that can tell you, for example, about January 6th, and Trump winning the election later, and Ice raiding those Chicago apartments in the dead if night is ~100M-10B. Individual runnable instances.

Go ahead, poison one. Someone's got the previous version.

Fascists will never be able to rewrite history again.

The current version of History will remain history, effectively, as long as computers/the internet.

Because of the massive amounts of data LLMs need, there will always be garbage input. So the overall quality is limited to "average" writing. Punctuation, grammar, and spelling can be imitated rather easily. Quality prose cannot.

On the flip side, imo there are some brilliant people with good ideas, and thier persuasiveness is hamstrung because they just can't stick to a word count that appeals to thier audience.

If you know that your audience has a 6th grade reading level... avoid a long form essay.

"When did neighborhoods become full of strangers?! We gotta get off our phones and get to know each other in our communities. We should be able to borrow a cup of sugar!"

Check it out, you can pitch socialism in a way that'll amp people up in like 3 simple sentences.

If you can't or won't do it yourself, by all means ask AI to help you lower your barrier for entry.

Basically all left (or even neoliberal) discourse happens in such an echo chamber people have forgotten how to talk irl.