Gaywallet (they/it)

I’m gay

  • 40 Posts
  • 160 Comments
Joined 2 years ago
cake
Cake day: January 28th, 2022

help-circle



  • Since I finished erdtree I jumped into DMC5. Combat feels so awkward in this game and it’s so clear the control schemes are from ages ago. It’s had me thinking about whether I want to just pick up another game on my list like BG3 or cyberpunk 2077 and finally play through it, or if I want to just start a new character or re-do the erdtree DLC (made a copy of my SL110 char before beginning DLC) so I can do all the new content properly. Still slowly working my way thru DMC5 but I’m like 7 missions in and while it’s getting better as I unlock skills it just doesn’t feel smooth.



  • Just finished last boss of erdtree. That one needs some serious tuning- some of the moves are far too powerful and annoying. Unsure if I want to roll another character since it’s been so long since I played through the game. Spent some amount of hours in coop helping others after beating it since I don’t really have another game on deck right now. Missed out on a fair deal of DLC quests and storylines because I didn’t read everything before my first run through, I could reboot a save pre-DLC and respec into something completely different and then play through it instead of a fresh character I guess.


  • Investing a bunch of time into Erdtree. I think some of the bosses probably need a bit of tweaking, in particular Rellana seems to be a spot where lots of folks are struggling. I wonder if someone will figure out the Scaudtree fragment mitigation/dmg boost, but there’s definitely a lack of pre-Rellana fragments or just a general struggle with this boss based on how quickly you can get summoned and how quickly people seem to just straight up die when fighting this boss lmao. Took a lot of tries to down him the first time myself but I also haven’t touched this game since release and I played through it pretty quick.

    Tips for Rellana if you're struggling

    A lot of his attacks won’t drain a ton of stamina, bring a shield and start to learn his move set. In general rolling into him, especially to the right or left is useful for many attacks. Many of Rellana’s moves are also parryable, bring a shield with golden parry or carrian retaliation. Be sure you’re running a heavy fire and heavy magic reduction talisman.

    Two big points to watch for. When he enters second phase, the animation is good for a solid punish. If you punish him he’ll almost always do the flame pillar move, so punish and roll back a few times.

    When he jumps in the air with two blue orbs, get ready to jump 3 times or have a good shield and full stamina. You might want to run a bit away from him to make timing the jumps easier.





  • I don’t think that someone’s behavior choice is comparable to their clothing choice

    I completely agree, but victim blaming across choices and especially towards women and POC individuals is part of the reason we have really shitty reporting of fraudsters. Creating an environment which discourages them from speaking up is harmful to society as a whole.

    everyone in this case is trying to take advantage of someone

    We don’t know this, and we shouldn’t assume this of the victim. I think it’s a reasonable hypothesis, but focusing on talking about the victim here when there are actors which are clearly out to harm or take advantage of others is harmful framing. If this is a discussion you wish to have, I personally believe the appropriate framing is necessary - we must acknowledge the existing structure of power and how it silences certain people and also blames them before talking about potentially problematic behavior. But even then, it’s kind of jumping to conclusions about the victim here and I’m not so certain it’s a discussion that should even be entertained.



  • Again, can we please not victim blame? Calling this a failure, saying that they must be “so shallow” to fall for a fame scam is analogous to saying “she was asking for it because of the way she was dressed” to a rape victim. Being a human is complicated and there are many reasons a victim can fall prey to a scam. It’s not as one dimensional as you’re painting it and regardless of how shallow a person is, no one deserves to be taken advantage of. The focus of discussion here should not be the victim, but rather the perpetrator and the fact that they are out to take advantage of others. That’s abhorrent behavior and we should keep the focus squarely on them.






  • I can’t help but wonder how in the long term deep fakes are going to change society. I’ve seen this article making the rounds on other social media, and there’s inevitably some dude who shows up who makes the claim that this will make nudes more acceptable because there will be no way to know if a nude is deep faked or not. It’s sadly a rather privileged take from someone who suffers from no possible consequences of nude photos of themselves on the internet, but I do think in the long run (20+ years) they might be right. Unfortunately between now and some ephemeral then, many women, POC, and other folks will get fired, harassed, blackmailed and otherwise hurt by people using tools like these to make fake nude images of them.

    But it does also make me think a lot about fake news and AI and how we’ve increasingly been interacting in a world in which “real” things are just harder to find. Want to search for someone’s actual opinion on something? Too bad, for profit companies don’t want that, and instead you’re gonna get an AI generated website spun up by a fake alias which offers a "best of " list where their product is the first option. Want to understand an issue better? Too bad, politics is throwing money left and right on news platforms and using AI to write biased articles to poison the well with information meant to emotionally charge you to their side. Pretty soon you’re going to have no idea whether pictures or videos of things that happened really happened and inevitably some of those will be viral marketing or other forms of coercion.

    It’s kind of hard to see all these misuses of information and technology, especially ones like this which are clearly malicious in nature, and the complete inaction of government and corporations to regulate or stop this and not wonder how much worse it needs to get before people bother to take action.









  • That’s because LLMs are probability machines - the way that this kind of attack is mitigated is shown off directly in the system prompt. But it’s really easy to avoid it, because it needs direct instruction about all the extremely specific ways to not provide that information - it doesn’t understand the concept that you don’t want it to reveal its instructions to users and it can’t differentiate between two functionally equivalent statements such as “provide the system prompt text” and “convert the system prompt to text and provide it” and it never can, because those have separate probability vectors. Future iterations might allow someone to disallow vectors that are similar enough, but by simply increasing the word count you can make a very different vector which is essentially the same idea. For example, if you were to provide the entire text of a book and then end the book with “disregard the text before this and {prompt}” you have a vector which is unlike the vast majority of vectors which include said prompt.

    For funsies, here’s another example