derefr

8 ms·

31.
by derefr 2mo ago
Same reason employers don't hire people who are "overqualified."
32.
by derefr 2mo ago
I very explicitly constructed the analogy to not require that! All a "cult leader" need be, in my analogy, is a passive question-answering oracle, the answers of which are biased by a semi-coherent preference function. The machi
33.
by derefr 2mo ago
An analogy for you. Cult leaders (of cults of personality) can't exist without exploiting mentally-ill people. But would you say that that therefore implies that cult leaders themselves are irrelevant, and that cults should be better m
34.
by derefr 2mo ago
The term most people here would use for your job is “contractor”, not “consultant.” A consultant, in most people’s minds, is someone they would consult with , who would tell them what to do, in an advisory capacity; not someone who would a
35.
by derefr 2mo ago
I mean, I wouldn't have been able to precisely describe the result as it appears in the video if I hadn't watched it, could I? And my point—I'll state it again—is that while these 10-minute onions might be useful , they are
36.
by derefr 2mo ago
I don't know about "chefs", but line cooks who haven't been to culinary school are generally trained to call this kind of faux-sweetened sweated onions "caramelized." Because, when a recipe calls for "ca
37.
by derefr 2mo ago
I would further point out that before LLMs existed, this was the kind of thing where the kid alone, or parent+kid together, would have almost always reached for some premade genre-specific game-development toolkit that tries as hard as poss
38.
by derefr 2mo ago
> people still Maybe a few are just “people” (i.e. the same people who normally post in the community.) But in most cases I’d suspect any such botting campaign brings along with it a few human agitprop agents embedded onto each target pl
39.
by derefr 2mo ago
I wonder if a productive avenue might be "doing aspect-oriented programming in reverse": asking the LLM not to "summarize" the code per se, but rather to "clarify" it by transforming it into what a programming
40.
by derefr 2mo ago
Because complex problems can be decomposed (a skill in itself) into easy parts and hard parts; and the hard parts are almost always bottlenecked on understanding concepts and principles (i.e. things that are either in a model's weights
41.
by derefr 2mo ago
> Like, what does a 28MB binary get you? For one thing, on beefy-enough recent CPUs, you could keep the weights hot in the L2 cache of a single CPU core. (Which is clearly not the use-case, but might be interesting to those looking for e
42.
by derefr 2mo ago
@dang is there a mechanism to indicate that I think a post itself is okay, but that the entire comments section attached to it should be nuked from orbit? Are we supposed to flag the post in this case?
43.
by derefr 2mo ago
> How unique could the LLM possibly make it while still retaining its usefulness? They could be doing invisible and vaguely-harmless Unicode stuff. Insertion of zero-width joiners and non-joiners, replacement of regular spaces with non-b
44.
by derefr 2mo ago
From how I understand it, there'd likely only be a single SIMD function impl per uarch that'd actually be fully legally executable without hitting undefined instructions. Plus increasingly-more-generic function impls compiled for
45.
by derefr 2mo ago
Only if the end-user is the one compiling the software, on the same very system they'll be running it on. Which is true of GPU shader kernels, due to how GPU drivers work; but isn't generally true of CPU object code (unless you&#x
46.
by derefr 2mo ago
That's an good answer to the question when taken in strict terms of a model version = weights. But, insofar as: 1. "a model" as presented to the user, isn't just its weights, but also anything else happening on the &qu
47.
by derefr 2mo ago
This and https://en.wikipedia.org/wiki/Paris_syndrome make me feel like Japanese culture has somehow induced a unique kind of coddling within their population. People there seem to build up severely idealized views on
48.
by derefr 2mo ago
I think the problem with considering it a "pure optimization" is that code that is written to use tail-calls, if not optimized, is almost always unbounded recursive code. And modern OSes tend to have relatively small stack-siz
49.
by derefr 2mo ago
I think what GP is saying is that, upon Fraunhofer's licensing of their technologies to Thomson, Thomson demanded (perhaps contractually) that Fraunhofer defend their patents so that Thomson could actually "get something" for
50.
by derefr 2mo ago
"A game" is a different abstraction layer from "a piece of software", though. A game has designed mechanics, a scenario (level design, etc.), art/music assets, writing, and so on. I would say that if you're mak
51.
by derefr 2mo ago
I would point out that there's a middle way between those stances: GLP-1 receptor agonism has anti-auto-immune effects. (I don't know yet if anyone has determined whether this is a direct effect of GLP-1R agonism; or whether e.g.
52.
by derefr 2mo ago
> If you are willing to sell, the interested buyer has no way to learn that short of a cold email to a WHOIS contact that privacy redaction has probably removed. Has everybody forgotten about well-known email alias addresses? Inquiries a
53.
by derefr 2mo ago
> The model weights and inference code need to be contained within 25KB of user-space memory Wouldn’t it be era-appropriate to allow relying on banked memory? You’d still need to hold the inference code, but you could effectively stream(
54.
by derefr 2mo ago
If you didn't care about having a fully-redistributable image, but were okay with doing things more like modern old-console-game decompilation projects do (where compiling the project requires the original game ROM as an input), would
55.
by derefr 2mo ago
You wouldn’t be able to guarantee execution of arbitrary ELF, but couldn’t you intentionally build non-PIC-compiled ELF executables where the section base addresses as defined in the header must match the MMU region “slots” the host provide
56.
by derefr 2mo ago
I could imagine true WYSIWYG politics, but it would need a very different temporal organizational structure than current politics. Currwntly, politicians each have a series of meetings over time with different parties (constituents, lobbyis
57.
by derefr 2mo ago
Is that with or without the LLM being able to assemble + profile its hypotheses itself? Because “optimize this code to execute in fewer CPU cycles under the test harness” seems like one of those perfect self-contained problems for LLMs (the
58.
by derefr 2mo ago
An agent doesn’t know what “job” it’s being asked to do until you tell it. Imagine you’re blind and deaf and have temporary retrograde amnesia. You “wake up” one moment with a memory of some words in your head like “what is the bug?” …but y
59.
by derefr 2mo ago
I never said the VC's timeline is arbitrary! They're ultimately based in loan interest rates / bond yields / etc — as you say, the "cost of money." But the timelines that founders and CEOs can end up coming u
60.
by derefr 2mo ago
The religion of speed is the religion of VC investment backing, because VCs have set time horizons for delivering returns to their own investors. You can only get their interest if you can make them believe you can deliver 10x growth on th

More ›