anuramat

6 ms·

61.
by anuramat 4mo ago
I was trying to be funny
62.
by anuramat 4mo ago
so, twitter went from 98% shit to 99% shit, and landfills get less smartphones; I'd say it's worth it
63.
by anuramat 4mo ago
> AI Agents experts you mean loop engineers?
64.
by anuramat 4mo ago
> NFTs and the Metaverse AI is not that bad
65.
by anuramat 4mo ago
you might want to read this post by a GA researcher: https://terathon.com/blog/poor-foundations-ga.html especially the part about duals -- made me feel like I was going crazy when I was trying to figure out degenerate
66.
by anuramat 4mo ago
the model did what it wasn't instructed to do by the attacker -- the "prompt" has basically nothing to do with the output
67.
by anuramat 4mo ago
> that's not fixing the problem, that's mitigating the problem is there anything humanity ever "fixed" then? surely it's possible in principle to solve at least some things that weren't solved yet > app
68.
by anuramat 4mo ago
don't want to be that guy, but have you tried nix?
69.
by anuramat 4mo ago
3) do you really not see how UAT is relevant to existence of a model with given properties? 6) so you think an OS is somehow a subsystem of software running on top of it? I'm kinda tired of this; you were mostly not wrong in the beginn
70.
by anuramat 4mo ago
> yes it is agree to disagree > every LLM has been vulnerable and every OS had bugs > show the logic https://arxiv.org/pdf/1912.10077 > you are the one asserting mappings existed I know? that's why I&#
71.
by anuramat 4mo ago
are you under the impression that LLMs and operating systems fill the same purpose?
72.
by anuramat 4mo ago
are all storage devices broken by design as well?
73.
by anuramat 4mo ago
the sound is forever etched in my mind; right next to ie with 12 bars, winamp and #008080
74.
by anuramat 4mo ago
yes, assuming bugs don't exist
75.
by anuramat 4mo ago
> clearly nothing ... is required this isn't even prompt injection; even if it was, how do you go from "exists" to "for all"? > we don't know the desired output then what are we talking about? if you don&
76.
by anuramat 4mo ago
> non-zero is it ever zero? is non-zero even a problem for sane usecases? > Dario are you saying claude reproduces CSAM from the training set? like, in ascii?
77.
by anuramat 4mo ago
but it's all just bytes?
78.
by anuramat 4mo ago
so, SQL injections and buffer overflows aren't unfixable because they never happen assuming nobody ever makes mistakes? under the same assumption you can just train your model until the output is correct
79.
by anuramat 4mo ago
> tweak badness enough assuming you get to do gradient descent AND the context is fixed+known AND you have unlimited compute? sure; is it a realistic setup? > the only way to fix ... the exact same argument applies to any (sufficientl
80.
by anuramat 4mo ago
> issues like prompt injection are unfixable how is it unfixable? do you mean "there's always a positive chance"?
81.
by anuramat 4mo ago
huh, I felt like (gpt5.5 < opus4.8 << fable) in terms of code quality, and (g ~= o < f) in terms of pure pass rate; which one did you mean? curious about your typical workflow/tasks I want to like gpt5.5 but it's like
82.
by anuramat 4mo ago
what you're describing doesn't sound like such a big deal -- it's (A) obvious during review, (B) easy to fix in a single prompt, (C) simple enough to fix manually, (D) can be mitigated with tokenmaxxing (agent review passes,
83.
by anuramat 4mo ago
> unlimited tokens for $50 a month link? > Why imho everything but opus produces unusable code (fable was even better...), eg gpt5.5 seems to write the absolute worst code that still technically solves the problem; tbh I'd be tot
84.
by anuramat 4mo ago
> dev tool do you expect people to quit research for some reason?
85.
by anuramat 4mo ago
I might at some point be spending more money on somebody elses claude subscriptions than on food my point is that the amount of calories a person needs is limited, and the efficiency is non-decreasing, so the per capita spending has an uppe
86.
by anuramat 4mo ago
isn't there a pretty big gap between a segfault and an rce? I thought that was the entire point -- that mythos closed the gap
87.
by anuramat 4mo ago
is nonzero leak rate sufficient for someone to practically exploit it? if you have to spend $10000 in tokens to get it to do what you want, is it still worth it? what if they manually review the requests of the users that trigger the guardr
88.
by anuramat 4mo ago
should we also restrict a model if it can clone a repo, set up the tooling and build a project?
89.
by anuramat 4mo ago
do you spend most of your money on food?
90.
by anuramat 4mo ago
> please don't explain why are you even talking to me?

More ›