Chu4eeno

12 ms·

241.
by Chu4eeno 5mo ago
It wouldn't have been that out of left field, he did work on massively parallel machines at Connection Machine. Though I guess that was more AI than distributed systems, iirc.
242.
by Chu4eeno 5mo ago
Yeah, Claude, Grok and Gemini immediately recognized it as a SUBLEQ VM (and worked out from there all the details), only chatgpt got stumped and just said it was a "visual simulator". All the big labs have plenty of proprietary (i
243.
by Chu4eeno 5mo ago
You know what happened at Google after Operation Aurora and they went full bore on security (BeyondCorp and all that)? They started phasing out Windows laptops for employees immediately. I'm honestly having trouble taking you seriously
244.
by Chu4eeno 5mo ago
This was obviously a conscious choice from the leadership at he frontier labs, and especially OpenAI, considering how 4o turned out. I don't think they expected the ELIZA effect [0] to explode as much as it did when they started includ
245.
by Chu4eeno 5mo ago
Well, it seems obvious (in hindsight) that attention is vital to actually do anything useful with all those matrices getting smashed together in increasingly unhinged ways in more and more layers. Which is incidentally more or less the only
246.
by Chu4eeno 5mo ago
It's not very novel, though, it's a fairly obvious step once you get something that can operate iteratively and largely independent, there were a ton of people trying to get LLMs to loop on their own even before deepseek r1. And I
247.
by Chu4eeno 5mo ago
Didn't neuroscience some big scandals about bad statistics and overstating their findings (in addition to normal issues like replication)? Look up at least the "dead salmon study" (hint: it's related to fMRI, and you can
248.
by Chu4eeno 5mo ago
I'm not a fan of him, but I'm still glad he picked up this. Just take a gander on the recent issues listed here, and how completely unavoidable they are: https://github.com/vim/vim/security I hoped the p
249.
by Chu4eeno 5mo ago
I only noticed it when a slew of security holes (re-)introduced by uncritical merging of LLM code got CVEs assigned. Luckily I was lazy running a custom build of vim and hadn't updated since they started lowering the merge barrier (or
250.
by Chu4eeno 5mo ago
I remember finding lcamtuf way back in the day because of that cheap robot guide he had. He lured me in with that, and got me interested in linux (and security) which I struggle with to this day. (Unfortunately it seems like he has taken ro
251.
by Chu4eeno 5mo ago
That's usually only necessary for more questionable things that was niche for a reason, like file transfer, images, audio, and rediscovering the root cause of old security issues like this all over again: https://web.archive
252.
by Chu4eeno 5mo ago
Just tested google's (lyria, integrated into gemini), and it made an honestly not bad progressive death metal song with female vocals alternating growling and melodic (though I accidentally used gemini pro to forward the prompt to the
253.
by Chu4eeno 5mo ago
Unique data like that is unlikely to have any impact on the learned/final weights after training. SGD, Adam and the other hillclimbing solvers abhor jagged edges from "novel" trade secrets and the like. Unless it turns out ev
254.
by Chu4eeno 5mo ago
You should just hit retry, usually they either latch onto the correct part of the latent space (surprisingly often) or they admit they don't know (or call a tool, depending on the model).
255.
by Chu4eeno 5mo ago
They are utterly trivial tools, and most of them are extremely over-engineered to the point of absurdity (that was the most embarassing about the Claude Code source map leak imho). That the for loop concatenating LLM output from POST calls
256.
by Chu4eeno 5mo ago
I think you forget that they really are stochastic (I'd wager it has hallucinated things to you that wasn't important so you missed it, or you've just been very lucky), and the people you're arguing with are forgetting t
257.
by Chu4eeno 5mo ago
Maybe try using them differently (I tend to use them like static analyzers I can yell at/argue with, and honestly less straining than trying to parse a Coverity report), or just avoid them. Mental health is more important than 20% gain
258.
by Chu4eeno 5mo ago
And Linux maintainers are actively pushing to radically cut down on the LOC by eliminating drivers etc.
259.
by Chu4eeno 5mo ago
If you can see enough of what an intelligence apparatus is doing to be impressed I'd say they're not doing a very good job. What they seem to be most unique in is the impunity of their operations, like when they assassinated a com
260.
by Chu4eeno 5mo ago
It has been for decades at this point. thiago's blog posts which introduced me to the topic over a decade ago (and is still one of the best explainers) points out that posix_spawn was introduced in POSIX.1-2001: https://web.
261.
by Chu4eeno 5mo ago
> It might be a "first-world problem", but having an account lost without appeal can justly be labeled "traumatic", especially if post-COVID it represents a majority of your social (or para-social) life. Also possibly
262.
by Chu4eeno 5mo ago
Yes, it's part of the GDPR (which really seems one of the most well-intentioned pieces of legislation once you read through it).
263.
by Chu4eeno 5mo ago
You need to hit the retry/regenerate button more, it's there for a reason. While the "stochastic parrots" thing is a bit overblown, IME most LLMs tend to surprisingly different responses even without changing the context
264.
by Chu4eeno 5mo ago
Isn't Pangram more expensive per token than most (all?) LLMs? Sounds too easily DoSable.
265.
by Chu4eeno 5mo ago
Wow, you weren't kidding, I found these before I got bored (they all seem to be LLM-authored from a quick skim): - https://github.com/godotengine/godot/pull/115280 Implement C# .NET Integration via Head
266.
by Chu4eeno 5mo ago
> I already acknowledged that "caste" is an incorrect word choice and I could've done better there, but my core point remains unchanged. Considering the reactions, it was a very amusing "mistake" to make (a bikes
267.
by Chu4eeno 5mo ago
> Their biggest products by revenue (search/adwords) and biggest stock value driver (AI/Gemini/Datacenters) are clearly in-house creations. ... or through acquisitions of doubleclick and deepmind?
268.
by Chu4eeno 5mo ago
Sorry for necro-replying, but assuming you're talking about cloudflare they already do. They took down KF (which is hosted in the US, so seemingly to be legal), but have always allowed everything from sites dedicated to livestreaming a
269.
by Chu4eeno 5mo ago
They should've social media outright if that was the point (it's no more healthy for adults).
270.
by Chu4eeno 6mo ago
Why do you view the judges as some superhuman unassailable truth oracles? Your view is extremely dangerous, tbh., and rather anti-democratic.

More ›