• 9 Posts
  • 846 Comments
Joined 3 years ago
cake
Cake day: August 4th, 2023

help-circle




  • OpenAI and Anthropic and Alphabet etc try their best to preventing it from doing so, so the act of getting it to do so anyway is “jailbreaking”, but yes. LLMs can reproduce works it was trained on. (At least chunk-by-chunk each limited to the token limit.) And I’d imagine the same is true of things like Stable Diffusion, though it might be much easier to get Stable Diffusion to produce something similar enough to qualify as a “derivative work” than it would be to get it to produce exactly the original.









  • I’ve got this trick. Basically, I start relaxing the muscles on the way to the restroom. Not enough that there’s any chance I’ll piss myself before I get there. Just enough that by the time I’m in front of the urinal, it’s not difficult to start the stream quickly.

    There was this one time I felt kindof bad about. went to the restroom where I worked (back before I was working remotely) and used “the trick” as usual. I get there, and there’s a guy at one of the urinals already. I peed, flushed, washed my hands, and left the restroom before he produced a drop.

    Of course, in all likelihood, “the trick” won’t continue to work so well as my beard grows grayer, and some day I’ll be the other guy.


  • Sometimes we purposefully make computers do nondeterministic things on purpose. Hardware random number generators, for instance.

    But not counting those situations, speedrunners often call some of the “other factors that just absolutely cannot be accounted for” kind of stuff that rethnor is talking about “RNG”.

    But even aside from that, there are some even less deterministic cases. Also from speedrunning, there’s the famous SM64 upwarp that my have been, but probably wasn’t a cosmic ray, but very likely was some kind of hardware failure that wouldn’t be accounted for by only considering how the N64 is supposed to work (given processor architecture and such).

    But an example I find even more interesting. There was a Defcon presentation where someone registered domain names that were one bit different from very frequently used domain names, kindof like typosquatting, but these were for domain names that no human was typing. These were domain names that computer programs were hitting, and the domain names weren’t derived from user input. But his one-bit-off domain names got a significant amount of traffic. Just from random memory corruptions and such due to some variety of weird hardware failures.

    I’d be skeptical of anyone I know claiming that they’d actually personally encountered something they knew to be truly non-deterministic in hardware that wasn’t in some way temporarily or permanently faulty or damaged in some way, but I think it is a thing that can happen in computers. And some of that nondeterminism comes from sources that actually do act at quantum scales were true nondeterminism isn’t unlikely to have sway.




  • TootSweet@lemmy.worldtoLemmy Shitpost@lemmy.worldMister!
    link
    fedilink
    English
    arrow-up
    5
    ·
    17 days ago

    Hey you, sit down and listen.

    Don’t be flippant and don’t be dismissin’.

    Think you’re a Flash encyclopedia?

    Eating, breathing Macromedia.

    Think you’re cool saying “all your base”?

    Get that Xiao Xiao out of my face.

    You gotta be kidding me with that crap.

    Animutation’s where it’s at.



  • TootSweet@lemmy.worldtoGames@lemmy.world*Permanently Deleted*
    link
    fedilink
    English
    arrow-up
    6
    arrow-down
    1
    ·
    27 days ago

    All I know is I’m not getting a Switch 2 until I can homebrew it. (Just as I didn’t get a Switch 1 until Fusée Gelée (the vulnerability in the Switch’s secureboot process that first enabled unsigned/homebrew code to be run) was discovered and made public.) Once unsigned code is possible, disabling region locking (assuming it’s enabled in the first place) would probably follow pretty quickly thereafter.


  • Well, to say “the voltage differential between A and B is positive” is saying exactly the same thing as saying “the voltage differential between B and A is negative”.

    There is such a thing as voltage differential. And the voltage potential at one electrode can be higher than the other, or the voltage at one can be lower than the other. But given that “voltage” isn’t really a thing except in relation to something else, there’s kindof no such thing as “negative voltage” or “positive voltage”. Just voltage relative to some other electrode.

    And if the voltage of one line is higher than the voltage of a reference (“ground”), then the voltage is said to be “positive”. If it’s lower, it’s said to be “negative”.

    I don’t think “negative voltage” is any less a thing than “positive voltage”, really, but neither one is really a thing except in relation to a reference. Which is to say you can’t connect a single electrode to an instrument and measure the voltage of that single electrode. That would be meaningless to try. You can only measure the voltage of an electrode by connecting your instrument the electrode and a reference.

    Hopefully that clears it up.