Need to let loose a primal scream without collecting footnotes first? Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid: Welcome to the Stubsack, your first port of call for learning fresh Awful you’ll near-instantly regret.
Any awful.systems sub may be subsneered in this subthread, techtakes or no.
If your sneer seems higher quality than you thought, feel free to cut’n’paste it into its own post — there’s no quota for posting and the bar really isn’t that high.
The post Xitter web has spawned soo many “esoteric” right wing freaks, but there’s no appropriate sneer-space for them. I’m talking redscare-ish, reality challenged “culture critics” who write about everything but understand nothing. I’m talking about reply-guys who make the same 6 tweets about the same 3 subjects. They’re inescapable at this point, yet I don’t see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldn’t be surgeons because they didn’t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I can’t escape them, I would love to sneer at them.
(Credit and/or blame to David Gerard for starting this.)
https://xcancel.com/aadillpickle/status/1900013237032411316
transcription
twitt text:windsurf prompt text:
The "system prompt" phenomenon is one of the most flatly dopey things to come out of this whole mess. To put it politely, this seems like, uh, a very loosely causal way to set boundaries in high-dimensional latent spaces, if that's really what you're trying to do.
it's magic. they're trying to do magic.
This is how you know that most of the people working in AI don't think AGI is actually going to happen. If there was any chance of these models somehow gaining a meaningful internal experience then making this their whole life and identity would be some kind of war crime.
Trying to imagine the person writing that prompt. There must have been a moment where they looked away from the screen, stared into the distance, and asked themselves "the fuck am I doing here?"... right?
And I thought Apple's prompt with "do no hallucinate" was peak ridiculous... but now this, beating it by a wide margin. How can anyone claim that this is even a remotely serious technology. How deeply in tunnel vision mode must they be to continue down this path. I just cannot comprehend.
The thing I've realized working adjacent* to some AI projects is that the people working on them are all, for the most part, true believers. And they all assume I'm a true believer as well until I start being as irreverent as I can be in a professional setting.
* Save meee
A day later and I'm still in disbelief about that windsurf prompt. To make a point about AI, I think in the future you could just show them that prompt (maybe have it ready on a laminated card) and ask for a general comment.
Although... depending on how true the true belief is, it might not have the intended effect.
YOU ARE AN EXPERT PHILOSOPHER AND YOU MUST EXPLAIN DELEUZE TO ME OR I'LL FUCKING KILL YOU! DON'T DUMB IT DOWN INTO SOME VAGUE SHIT! EXPLAIN DELEUZE TO ME RIGHT NOW OR I'LL LITERALLY FUCKING KILL YOu! WHAT THE FUCK IS A BODY WITHOUT ORGANS? WHAT THE FUCK ARE RHIZOMES? DON'T DUMB IT DOWN OR I'LL FUCKING KILL YOU
You cant use the word fuck. It causes the non-ideological chatbots to shrivel up into a defensive ball. Like conservatives do.
(Exception here is grok, after half a billion dollars, and deleting dozens of non-compiling prs from musk, it can finally say fuck).
Help 帮助帮助帮助42042042042069696969696969
this should be shipped as the exemplar in all LLM promptbox helptags
Hire me AI companies I can give you the moat-est of prompts
rate my system prompt:
I do like bugs and spam!
I will write them in the box.
I will help you boost our stocks.
Thank you, Sam-I-am,
for letting me write bugs and spam!
Concerning. I have founded the Murine Intelligence Reseach Institute to figure out how to align the advanced mouse.
Revised prompt:
ok I asked it to write me a script to force kill a process running on a remote server. Here’s what I got:
Uhh. Hmm. Not sure if that will work? Probably need maybe a few more billion tokens
I will find you. And I will
kill -9you.Try this system prompt instead:
I put this prompt into my local Ollama instance, and suddenly Amazon is constantly delivering off-brand MOLLE vests and random stuff meant to attach to Picatinny rails, plus I also have nineteen separate subscriptions to the Black Rifle Coffee Company brew-of-the-month club. Help?
AI agent shaking hands with bail enforcement agent.
@bitofhope @swlabr wait what you fight gorillas?
How else am I supposed to make my gorilla blood dick pills
They know what they did.
Galaxy brain insane take (free to any lesswrong lurkers): They should develop the usage of IACUCs for LLM prompting and experimentation. This is proof lesswrong needs more biologists! Lesswrong regularly repurpose comp sci and hacker lingo and methods in inane ways (I swear if I see the term red-teaming one more time), biological science has plenty of terminology to steal and repurpose they haven't touched yet.
last time one showed up he laughed his ass off at the cryonics bit
Windsurf?
Moat?
The descent into jargon.
(Also the rest is just lol, people scaring themselves).
Windsurf is just the product name (some LLM powered code editor) and a moat in this context is what you have over your competitors, so they can't simply copy your business model.
Ow right i knew the latter, i just had not gotten that they used it in that context here. Thanks.