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.

Is this AI generated?

Reject internet

Return to Ham Radios

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.

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.

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.

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.

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

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.