Just two weeks til’ Halloween now.
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.)
New Odium Symposium premium episode. We talk about Julia Galef guest hosting the Ezra Klein show to shill superforecasting
https://www.patreon.com/OdiumSymposium/posts/ezra-klein-is-2-172872571
Quick Q, from where does the weird term “abliteration” come from? I’ve seen it used for “jailbroken” GenAI models.
With a bit of google searching by date it seems like this guy (https://huggingface.co/failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5) claims ownership of the coining, which I find plausible, based on the list of released weights and their timing.
As far as “abliteration”: it’s just a fun play-on-words using the original “ablation” term used in the original paper to refer to removing features, which I made up particularly to differentiate the model from “uncensored” fine-tunes. Ablate + obliterated = Abliterated.
The paper being referred to being: Refusal in Language Models Is Mediated by a Single Direction which seems to have somehow started out on the “AI alignement forum” of LessWrong https://www.alignmentforum.org/posts/jGuXSZgv6qfdhMCuJ/refusal-in-llms-is-mediated-by-a-single-direction first posted 27th of April 2024. In general it’s a method of weight alteration so only really applicable to some opensource (or i guess leaked) models.
First model with abliterated in the name by “failspy” on 8th of May 2024.
Thanks!
wiktionary claims it is a blend of “ablate” and “obliterate”, it does not list an origin (at this point in time)
edit: looking a little more into this, knowing the etymology requires reading about AI discussions which I am not prepared for at this point in time but it is out there in the internet
New OpenBSD is out and the commemoratove artwork is definitely trying to say something about genAI.
On a more personal note and trying to keep exact circumstances vague and not name any names, but the IT corp I work at just had another wave of layoffs and our team got reduced from 5 to 4 people (and same for a neighbouring team that we are pretty close socially with, and more). Idk how they expect us to get anything done (I mean, I know, they want us to use more AI), especially with how much the mood has tanked. By any metric productivity this week is going to be close to 0% because everyone from affected teams has just mentally checked out.
So uhhhh anyone has any recommendations for what fields I could try to retrain to from IT? Preferably without going through uni again. I’m thinking locksmithing.
If you’re interested in locksmithing I’ve got bad news. It got harder after WFH, waaaay more people home in rich neighbourhoods
OT: popping in for a second since I’m not drowning in homework for once (midterms start friday) to say carpal tunnel stripping is really good.
(so far I’m averaging in the high 80s last I checked).
Good luck with your exams.
going thru it myself, can report that non-graded “suggested exercises,” with exams being the only graded material, is not any better
So I just stumbled into Team Human, because a Youtuber I kind of like (Kraut) published a video for them. They seem to run a campaign targeted at creators, pushing to get videos released on the 10th or 17th. So they must be having a bit of a high point right now. (Idk if cash changed hands.)
Their manifesto tries to amp up societal harms, but the existential risk is snuck in at the end (AI programmed pandemics, etc.) and they want to regulate AI chips like nukes. CAIS (Center for AI Safety) is listed as the sponsor. Seems a little “AI centrist” to me, as in aligned with the less insane billionaires, but still a typical AI safety astroturfing shop. Shame if any of your favorite creators fell for it.
Glenn Beck is listed as one of the “leaders”. That guy is still around?
Iirc he has been trying a ‘im no longer a frothing weird rightwinger’ for a while now, but so far nobody really fell for it.
People who are entertaining in front of a camera are just regular people. If anything I’d expect the fame to make them on average worse than a random sample
They got one of my favourites, Medhi Sadaghdar (ElectroBoom), and yes in a video released on the 10th.
CAIS seem like an “AI safety” laundromat, in the sense they try to put a reasonable face on the paperclip maximiser stuff. They were the ones who organised the “AI safety statement” in 2023, which features lots of signatories who I’m sure just thought “oh yes I don’t want anything bad to happen because of AI”, not “If we don’t give Yudkowsky the reins we’re doomed to become office supplies”.
Not to be confused with Team Human the Douglas Rushkoff podcast.
My first association was Alex Jones screaming “I’M TEAM HUMAN” while ranting about environmentalists. So the brand is getting a little crowded it seems.
I was gonna ask about that
Decent monday morning take from robert evans:
Schrodinger’s double-think.
transcription
Quote skeet:
This is a researcher at Anthropic. Some of these people have lost their minds! What is mainlining Claude all day doing to them?
there are basically two possible futures at this point. either we’re all going to die of[sic] the future is going to be Good beyond our wildest imagination. the trick is to hold this duality in your head and not be driven insane by one side.
Robert Evans (the only Robert Evans):
some of these people get so locked into a labyrinth of Rationalist thought experiments that they are no longer really capable of interacting with reality 1/2
you can in fact limit the horizons of possibility in your own mind so severely that you are no longer able to “think” in a useful way 2/2
Well, we are all going to die. So… musing over?
We all get that right, everyone is going to die. Everything living you lay eyes on is going to die. I am 100% certain that the future contains the death of everyone alive today.
Do they mean immanently? Why?
They’re referencing the meme that an uncontrolled AI will inevitably kill every last human being, for reasons.
They seemed to bring up a cartoon for 12 year old boys?
Not if you’re a Yudkowskian rationalist. He’d call us “deathists”.
I don’t have to be thrilled about death to have a basic understanding of how basic our understanding of consciousness, biology, and neurology are. Nor to know how much the physical world resists making tiny machines and measurements necessary to explore these frontiers, or the ethical dilemmas involved in exploring how consciousness works that vastly slow and perhaps outright prevent answering questions like “can we replace little bits of someone with some machine we can maintain easier in the interest of prolonging them?”
Also I do happen to know how the universe ends.
I don’t have to be thrilled about death to have a basic understanding of how basic our understanding of consciousness, biology, and neurology are.
I think its the mid-wit meme, lesswrongers know just enough to imagine some grand solution like brain uploading or nanobots might cure death and not enough to appreciate all the reasons each of those solutions won’t work.
One of the funniest things I ever read was a result of a survey of laypeople, physics UGs, physics Grads, postgrads, and faculty on various “cool future” things.
predictions of the likelihood of things geeks constantly dream about craters once you get past the very basics. I can only assume that fields that aren’t mine, and are in general vastly more complex, are at least as full of evidence of god hating us and not allowing us nice things.
I remember reading some of yuddies writing about nanomagic and noticing that the word “sticktion” never showed up.
Judging by the way LLMs and generative AI has been used, even when we do get something vaguely like a realistic version of sci-fi fantasies, capitalism and people who can’t distinguish the limits of the real thing from the fantasy will manage to misuse it.
This doesnt apply to everything, like VR and 3d printers were a sci-fi fantasy and now they are mundane economically viable but not revolutionary products.
Biohacking and nanomachines otoh dont seem to be viable and you hear nothing about those anymore. Only a little bit in the uncreative (and going on the experts unrealistic) ‘ai will create bioweapons’ critihype. No ‘It will turn people into cyberzombies’ (quake 2 was pretty good (that was aliens not AI but still)).
who in the hell would want programmable matter anymore, after they’ve seen what happened to plain ol’ computer programming as already practiced
Piles of screaming grey goo and people being like “Why are you running away? can’t you see it’s my grandmother?”
What’s that XKCD that goes over the ways that capitalists would be ruthlessly exploiting all those “secrets they don’t want you to know” for incredible profits?
Ah yes ‘only these two things are possible, dont consider any other outcome’ truly rational thought
Also, ‘it’ll be Good beyond our wildest imagination’ is just the vaguest thing ever. ‘yeah bro AI is gonna make everything super good. How? Why? uhhh i dont know it just will’
‘it’ll be Good beyond our wildest imagination’
pumpkin spice season will be permanent
AI will make free ice cream for everyone
That ice cream? Also pumpkin spice
We could unironically live in this world if only the powers that be would allow it. I’ve personally radicalized at least 3 basic bitches with this fact.
They’re also incredibly horny about creating conscious digital slaves for some godforsaken reason
(To the tune of the teenage mutant ninja turtles theme*): self-inflicted psychic debuff
*The correct theme, not the heretical one. If you have to ask, you’re wrong
It isnt Claude that did this to them, it is their desire to be lesswrong.
wat
Lobste.rs dropped the vibecoding tag because a prominent LLM enthusiast had a sad and huffed off in a fit.
It’s unfortunately common that authors see their work posted, join the site, but then only use us for traffic. Hopefully this time helps them understand the difference between a symbiotic and parasitic relationship.
This feels like a very weird thing to say. (I’m reminded of some weird things Elon Musks mom said about how twitter users should be more grateful for all the traffic or something).
I mean the bit above it makes sense (making a hard only ‘25% of post can be self promotion’ seems like a recipe for disaster however) but this added line made me raise a few eyebrows.
I really dislike how he believes that LLMs and fascism are “separate axes”. Where does he think the LLMs come from? Does the stork deliver them? Is ChatGPT powered by magic fairy dust? Does he think that Sam Altman is a great defender of liberalism and democracy who definitely hasn’t donated tons of money to Donald Trump?
I can’t help but think it’s another instance of the unquestioning Californian Ideology assumption that all new technology is inherently liberatory.
We try to confine that shit to SF
This can either be read as San Francisco or Science Fiction, and both readings work perfectly
It’s not working, I spotted a probable hackernews all the way out in Modesto last week, they’re like the goddamn nutria now
He works for google, so he probably still thinks they’re not evil.
Oh, buddy, big feelings
I am saying that the anti-fascist communities ended up […] uninterested in using LLMs
:O
Very disappointing, even if I was only ever a non-registered lurker. Gave in and DNS-blocked it and HN this morning (I had a HN-filtering userscript which in hindsight was just cope).
Lobste.rs is analogous to methadone treatment for Hacker News, prove me wrong
Now it’s not even good for that anymore.
I considered trying to get an account there, but the last ~18 months or so have made me feel the same way about Lobste.rs as I do about Bluesky. If you’re just running a lower-traffic version of the thing you’re trying to “improve” upon with slightly different moderation policies, you’re not going to build a lasting platform or community.
came here to post that. Genuinely having to reconsider being an active user there now
It’s been a day, and the mods are not sitting still:
I see you were one of the people whose comment got censored. I presume the censored comments were about the parody of the “pelican on a bike” animation on the OpenBSD 8.0 release page.
So now people can’t even discuss criticism of LLMs on lobsters? Insane.
Might as well start this off
so we all remember this graph from AI 2027 right? the graph that the whole paper relies on?
well its wrong. not just normal wrong, it’s ‘the authors literally disowned it’ wrong
(ignore the claude mythos purple x i took this from a reddit screenshot) but they completely changed their predictions in december 2025, several months after the original paper was published
Is this the thing where the baked an asymptote to infinity into their “model”? It was both bad modeling and also made all of their other input “data” completely irrelevant. I remember someone pointing this out at the time AI 2027 dropped, but there was so many other problems with it in the sneer spaces we didn’t talk about it too much among all the other problems. And someone on lesswrong actually pointed it out, but somehow, “we accidentally baked an asymptote into our model that is supposed to be detailed and carefully thought out” wasn’t a breaking point for lesswrongers.
Not literally an asymptote, but once you presuppose something as farcical as nested exponentials, your model is doomed to blow up immediately after the last data point.
I believe so
That ‘superexponential’ thing annoys me so much. Just the most blatant ‘the line goes up because I really really want it to’, while all their rationalisations for it are just describing exponential behaviour.
We talk about the importance of humanities but it’d be nice if they learnt a bit of maths too.
Not sure if the book also did super exp lines, but the graphs in the singularity is near also have various issues that never got to much deep thought.
Always found it funny that the book admits basically (no direct quote): “sure tech follows an S-curve normally, and that looks exponential at first, but if you take the rate of all historic new techs together on a timeline that clearly is exponential”.
And I just thought, you cant know if it is, thats the point with the S-curve as well, and it messes people up. (Like Disco Stew from the Simpsons), and your N=1, it might not even be a S-curve, it could just have a sudden ceiling.
Despite wanting to believe, I was less convinced. (also, not sure if it was just my copy but Yudkowsky’s name was spelled wrong, still looking for somebody who has a physical copy who is willing to check for me if that was also in print or just a pdf error)
Back to the subject, a course in business management would have also helped here.
“Sigmoids always add to an exponential curve!” is another standard futurist trope from the 1960s and 1970s. Sci fi authors used to share this diagram with a curve fitted to pony express to trains to automobiles to aircraft to chemical rockets. “Nuclear rocket” and “interstellar spacecraft” were helpfully sketched in as the obvious next steps. Needless to say, we did not develop drives which can reach 10^-3 C around the year 2000 and will not reach 10^-2 C (just 430 years to Alpha Centauri) by 2050.
this is fucking time cube lol
woah damn why does this look exactly like some Kurzweil bullshit
I have not read any Kurzweil for 20 years but I think if you read him most of it will be recycled '50s-'70s tropes. Even nanotech appeared in a Feynmann talk, and Moore’s Law is from 1965. Remember, cranks keep presenting ideas which were refuted long ago as new.
