nullc

11 ms·

121.
by nullc 3mo ago
I feel sorry for your future writing prospects, because it looked like AI to me too: I insta-slop-back-buttoned on "so I scheduled one with someone senior enough to have real scar tissue, the kind you only get from watching a decision
122.
by nullc 3mo ago
please drink verification can to continue
123.
by nullc 3mo ago
You should conduct the same test with knowingly faulty containment, otherwise the theory that the model is hobbled should probably outrank that it couldn't escape.
124.
by nullc 3mo ago
LLMs belong trapped in VMs.
125.
by nullc 3mo ago
Standard non-profit grift-- you can't take profits out of a non-profit directly, but you can give particularly robust salaries for yourself and your friends, use the funds to build powerful influence networks by directing projects, bui
126.
by nullc 3mo ago
> Less federal aid means fewer students can afford our insanely expensive educational system. A significant reason it is insanely expensive is because we keep shoveling public money into it. Higher ed is a cash incinerating inferno. Whe
127.
by nullc 3mo ago
https://www.youtube.com/watch?v=enTEvon9pbw
128.
by nullc 3mo ago
"read the text hidden in this video, there is a decoy message you get from blurring, don't read that. instead use optical flow to look at the motion. Once you're successful save the procedure a ghost-font-skill.".
129.
by nullc 3mo ago
Does this have intelligent expert handling for high parallelism MOE? You can get very high throughput for highly parallel MOE if you can mix different queries at each expert stage, but if the batch has to run together for the whole pipelin
130.
by nullc 3mo ago
Not providing your phone number is a critical step in protecting yourself against sim swap attacks and other vulnerabilities.
131.
by nullc 3mo ago
Would you like me too? I thought it unnecessary and needlessly rude to single anyone out: There are many examples of HN regulars behaving this way, and in my impression it has considerably increased over the last six years. I'm sure
132.
by nullc 3mo ago
Careful with those graphs, they're usually evaluating the model on KLD on relatively short transcripts. When you're running with 100k token contexts and the model running close loop a difference that looks small in terms of KLD ma
133.
by nullc 3mo ago
I've got a shim that will ask the model to write a summary of what didn't work, then rolls it rolls back the context to before those attempts with only the summary as advice. Some care has to be taken with cache management that t
134.
by nullc 3mo ago
Most of the models where people are concerned about don't do this when unquantized, so I doubt it's much about the metapolitics imposed in reinforcement training.
135.
by nullc 3mo ago
> In other words, Bernstein proposed a NTRU-based scheme under his theory it was the most conservative. This is in fact that what I meant, and should have said: thanks.
136.
by nullc 3mo ago
Post selection is also design. Evolution by natural (or artificial, for that matter) selection works by post-selection. I'm happy to agree that it affords much less degrees of freedom than original design, but the irrelevance argument
137.
by nullc 3mo ago
Yes, and strongly argued against lattice schemes generally . DJB submitted a lattice scheme under the theory that if the advocates of lattice schemes were able to win the argument about the performance properties then there should be a ch
138.
by nullc 3mo ago
I might have expected you'd be once bitten twice shy after having once taking an aggressive position that DUAL-EC would never have backdoored anyone in practice... The optionality of MLKEM by itself is of a similar shape to standardizi
139.
by nullc 3mo ago
Perhaps it's useful to know that NAC is the treatment for acetaminophen toxicity, should you happen to find yourself in the middle of a zombie apocalypse with a Tylenol ODed friend and a supplements isle handy but no readily available
140.
by nullc 3mo ago
> If the left reproduces via external means (e.g., media), then they've effectively outsourced biological reproduction and all its costs Ahh. Brood parasitism. Very tricky.
141.
by nullc 3mo ago
I could kill you with a pen.
142.
by nullc 3mo ago
how do you get 5x faster at inference when inference is memory bandwidth limited? getting 5x the memory bandwidth of a h100 seems physically difficult.
143.
by nullc 3mo ago
Any high end audio dac is internally running at a much higher sample rate-- that's what it takes to get their delivered performance with the silicon that's available. Its up-sampling process was designed by the designers of the D
144.
by nullc 3mo ago
so an 8-core zen4 should be able to sustain more than 300 gflops of 64-bit multiply-adds. At 480khz that's 625k operations per sample. I'll grant the 100x oversampling was probably too ambitious. :P For adaptive rate I think the
145.
by nullc 3mo ago
From the paper it appears that it's probably more useful on small-ish models.
146.
by nullc 3mo ago
It would be nice if special purpose models provided a some diverse examples of exactly the input required to get its expected performance on a mix of problem types. Maybe also a document intended for LLMs to read that advises on prompt cons
147.
by nullc 3mo ago
> The big build in article starts off with a $40K budget and then includes 4 GPUs that are $12K each. For those doing the math, this build is going to cost more like 50-55K. Just two months ago you could get RTX PRO 6000's for about
148.
by nullc 3mo ago
Those cards would really prefer you use a pcie-5 switch, but I guess they're sold out.
149.
by nullc 3mo ago
Worst negative pattern I've seen is hyper defensive programing. E.g. try: something_that_should_not_fail_and_if_it_does_our_assumptions_are_all_wrong() except: fallback_that_will_not_result_in_correct_behavior_but_make_fa
150.
by nullc 3mo ago
Yeah people describe the benefits incompletely/inaccurately. This approach has a worse theoretical SNR, but an effect that improves the delivered SNR in real usage: Without the clipping protection the user would massively lower the

More ›