vidarh

25 ms·

601.
by vidarh 6mo ago
It doesn't need to be easy, but unlike with encryption it also doesn't need to be particularly precise . E.g. it's okay to exclude non-poisoned training data because you didn't manage to create a precise enough detector
602.
by vidarh 6mo ago
Rice's theorem is a close corollary, but I did mean the halting problem. Pointing to the halting problem was a bit of a throwaway quip because the "general shape" of it is an easy smell test for whether something is likely to
603.
by vidarh 6mo ago
This is why I pointed out that the only way poisoning has a chance of working other than over very short timelines is if the tools to do so remains private and inaccessible to the public. It's a bit of a leap, but the halting problem c
604.
by vidarh 6mo ago
It's the very core of the halting problem.
605.
by vidarh 6mo ago
If they only poison something "nobody" cares about, then sure, nobody will care about it. In which case the poisoning also won't matter.
606.
by vidarh 6mo ago
> Why would they be in business if self play can work out so well? Because it is still cheaper.
607.
by vidarh 6mo ago
> the ability to poison models, if it can be made to work reliably Ultimately, it comes down to the halting problem: If there's a mechanism that can be used to alter the measured behaviour, then the system can change behaviour to t
608.
by vidarh 6mo ago
GLM 5.1 is the first model I've found good enough to spring for a subscription for other than Claude and Codex. It's not crushing Opus 4.5 in real-life use for me, but it's close enough to be near interchangeable with Sonne
609.
by vidarh 6mo ago
When Sonnet 4.6 was released, I switchmed my default from Opus to Sonnet because it was about en par with Opus 4.5. While 4.6 and 4.7 are "better", the leap is too small for most tasks for me to need it, and so reducing cost is no
610.
by vidarh 6mo ago
Mao's campaign to kill sparrows was a result of a belief that they were a net loss for harvests. Stalin's support of Lysenko was a result of thinking Lysenko was actually able to drive agricultural growth. Both mistakes led to mas
611.
by vidarh 6mo ago
I hadn't thought about whether this would still with modern speakers, but this was the common assumption for several older types of speakers and microphones. One of the first "science experiments" my dad showed me was the oth
612.
by vidarh 6mo ago
Same. I've used Rails a few times, but something like 95% of my Ruby use over the last 21 years has been non-Rails.
613.
by vidarh 6mo ago
It doesn't typically get run millions of times per day because in most regular uses it's trivial to cache the glyphs. I use it for my terminal, and it's not in the hot path at all for rendering, as its only run the first ti
614.
by vidarh 6mo ago
Memory savings and by extension memory bandwidth for low color modes, as you indirectly point out with the text example. The early Amigas had a memory bandwidth of ~7MB/s, and the chip RAM bandwidth was shared with the custom chips..
615.
by vidarh 6mo ago
> marketed as a 32 bit system, in 1985. I'm sure there must have been some, but most of Commodore's early Amiga ads didn't mention the number of bits at all, and from looking through old magazines it doesn't seem most
616.
by vidarh 6mo ago
Looking up the hash of a tree in git is few enough operations that I would be very surprised if that is true for all but the smallest caches. If you were to shell out to the git binary, maybe.
617.
by vidarh 6mo ago
For pretty much everything. My terminal is in Ruby, with a Ruby font renderer, running Ruby shell, and my editor is in Ruby, my window manager, my file manager. (Yes, I'm taking it a bit far; my prototype Ruby compiler is self-hosting
618.
by vidarh 6mo ago
From the article: "This new feature will be available in Ruby 4.1.0."
619.
by vidarh 6mo ago
> More importantly, on CI systems it’s relatively common to check out code using git, and git doesn’t care about mtime git doesn't care about mtime, but git maintains trees whose hash changes if any constituent part of the tree chan
620.
by vidarh 6mo ago
I assume they mean: https://github.com/mksglu/context-mode
621.
by vidarh 6mo ago
Yeah, I second that 16 bit or 16/32 was far more commonly used than 32, due to the 16 bit bus.
622.
by vidarh 6mo ago
Of course that is not at all what I wrote.
623.
by vidarh 6mo ago
In Norway, Danish is sometimes subjected to friendly ridicule as sounding like Norwegian spoken with a potato in your mouth... Danish is if anything ever so slightly closer to German in vocabulary and grammar, but the pronunciation is anoth
624.
by vidarh 6mo ago
Yes, coming from the elite strata is and has objectively been considered signifiers of high status and prestige in most societies throughout human history. I sense that you are trying to imply that I am applying some kind of value judgement
625.
by vidarh 6mo ago
> What I questioned was its truthfulness You can question the truthfulness all you want. It doesn't change the fact that this has been a well established aspect of the language struggle in Norway for well over a century, to the poin
626.
by vidarh 6mo ago
Or using the copper. Plenty of Amiga games played palette tricks despite not using HAM.
627.
by vidarh 6mo ago
HAM doesn't flicker. The issue with HAM is that you're limited in terms of abrupt colour changes. It's straightforward to convert HAM to PNG etc. It would have sometimes been used together with interlaced mode to double the n
628.
by vidarh 6mo ago
> What do you know about Bokmål being more prestigious in the past? Having grown up in Norway and seen first hand how it was treated that way. > You don’t respect the other form enough to cognize that it exists.[1] I don't like N
629.
by vidarh 6mo ago
Yeah, same thing for me. The article is funny, but I'd hate to actually use a language using non-English keywords.
630.
by vidarh 6mo ago
Word Basic back in the Word 95 days translated keywords . The source was saved tokenized, so the program would have different keywords when loaded in different version of Word. I don't know if there was a Nynorsk version, but I presum

More ›