• 1 Post
  • 6.99K Comments
Joined 3 年前
cake
Cake day: 2023年6月30日

help-circle





  • I’m gushing a bit about the games and that part of me thought they might be the best games I’ve ever played and the other part of me that would normally say hold up a second is having trouble coming up with any strong contenders. Subnautica maybe. Metroid Prime, if the controls were better. Hades would make the top list but wouldn’t top them. Ori rivals it for beauty and atmosphere but not controls and combat; it’s not even close there (though it’s possible I just need to get farther in to really appreciate Ori).


  • I just started that one, after finishing the first one and immediately jumping into my second play through. I’m not very far in so far but wow, it’s already looking like an amazing sequel for a first game I’d already describe as flawless. Like I got so used to the first one that the early game seemed almost trivial in that second play through but silksong’s enemies don’t follow the same patterns and are able to hit me pretty regularly. And both games are filled with this strange bleak charm.






  • Oh yeah, it’s absolutely understandable. But also frustrating because division tactics only need to point out things to be angry about and then people will get angry about them (might have already been angry about them but now the topic is on their mind) and yell or lecture people about it which feeds that division and makes it less likely they’ll ever get on the same page. Especially when the offending side was trying to act in good faith but gets accused of just being an enemy or bad person.

    Shame only works when the person decides on their own they should feel shame, otherwise the best you can hope for is platitudes while they secretly want to see you fall (or otherwise get out from under your sphere of influence) so they can stop pretending to care about that thing. Average case would be an argument that leads nowhere, which is what usually happens online since no one has any real power over others, other than mods/admins who can only ban.



  • Yeah, the LLM I asked also got it right when I pointed out the error, but I’m not trying to say that LLMs can’t get things right, but that they won’t ever be consistently right and that the wrong answers will look just like the right ones. As in if you know what you’re talking about, you have to catch the errors, and if you don’t know what you’re talking about, there’s no way to know whether the answer you just got is accurate or bullshit.

    Systems that rely on LLMs that don’t have a way of automatically verifying what the LLM outputs (and programming only partially applies for this) will fail randomly.

    Another example: at my job, we have a system that adds in special messages for the LLM when it uses hooks. One of the sub-agents became suspicious of these messages and reported to the main agent that something was injecting false data into its context because one message reported a date change and also had to say “don’t tell the user, they are already aware that the date has changed”. The original agent didn’t even clue in that they were the same messages it was seeing until I pushed back.

    Two instances of the same thing treated the same messages very differently and the one supposed to manage it all didn’t even notice until it was told. That’s the quality of these things. And it’s no wonder when the same data stream is used for actual data along with instructions (which is just data because it doesn’t take instructions, it predicts text and can look like it’s taking instructions because it predicts text based on a context that includes the instructions).