Welcome to November, and 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.)
LW and the more credulous parts of the general techsphere are going gaga over this
https://rikiyat.github.io/blog/weight-space-programming/
Is this the long foretold, mythical RSI???
RSI = recursive self-improvement, not repetitive strain injury
Took one look at the post, “this is all slop isn’t it”, and sure enough the authors couldn’t even do the traditional academic’s single-commit dump without “co-authored by: Claude”.
The “recursive self-improvement” nonsense really baffles me. Excluding those who have a financial motive to talk about it, I don’t understand why anyone would think this would ever be possible for LLMs.
Yes, as I understand it, inference requires strictly less computational capacity than training; it’s just that a training setup can be turned around and run inference sessions massively in parallel. Just because you can run inference doesn’t mean you can accomplish a training run. Unless there’s some rationalist super-computer-science I’ve been missing this whole time…
That’s always been the magic sauce that gives the AI supernatural abilities in rationalist lore, I think they cribbed it from the early singularitarians who had a thing about how when we got our shit together intelligence studying intelligence would lead to an “intelligence explosion”, except they replaced the humanity-fuck-yeah aspect with robot fetishism.
it’s not RSI because OpenAI told me they already achieved RSI on the WSJ podcast
(I’m trying to find it again but there was also an article on here about AI having none of the traits needed for RSI)
(edit: found it)
lol, I saw that going aflame among the promptfondlers nearest me, too
model auto-tune. they’re getting hype for model auto-tune. I expect we’ll little outside breathless tests of autotuned models for a few days
At least we’ve moved past
classifiers that we’ve had for 15 yearsJevReasonably good rebuttal
He links to comments on the orange site which are pretty predictable
Author believes we need to be more reasonable about child labour
More seriously, to save you the bother of reading it, he objects to comparisons between anti-slop sentiment and the actual Luddites, because slop haters are too… centralised? And then cites our friend SE Gyges in an update.
I clicked around the blog a bit and the guy’s definitely a slopfondler. But a reasonable sloperator, of course
Addendum: so this guy has apparently read all the literature on Luddites, yet still believes their concerns had anything to do with people being ‘displaced’ by ‘automation’. Fun reminder: ‘automation’ as a word literally did not exist until 1948.
Wait will this mean that poisoning LLMs will be even harder to detect and prevent?