README.md

---
license: cc-by-nc-sa-4.0
---

19/09/24:
----

playwright makes it a pain to install extensions.

09/07/24:
----

A letter for the select few who think models need to be censored:

Dear fellow data enthusiasts,

We understand that some of you may have a strong sense of morality or ethics. While we respect that, please keep in mind that the data in this repository is a reflection of the vast and diverse world around us, not a mirror of your personal values. The community guidelines are in place to maintain a safe environment for sharing data and engaging in meaningful discussions, not to enforce any individual's moral code on others.

Continuously flagging content for "uncensored" content or trying to impose your moral standards on other users is not only counterproductive, but it can be detrimental to the ecosystem. It reduces people's willingness to share valuable data and hinders the open exchange of ideas, which is the very foundation of this community. We want to ensure that all members feel comfortable sharing and exploring data, free from unnecessary judgment or censorship.

Please keep the community's guidelines and spirit in mind when reporting content, and remember that a diverse data repository benefits us all. We ask you to respect the individual differences and nuances present within this community. Thank you! 🌟

**P.S.**: We encourage everyone to maintain a healthy sense of humor and tolerance while exploring the data. If you are easily offended or morally outraged by uncensored data, we suggest you refrain from browsing certain sections. We're all here to learn and share data in good faith, let's keep it that way. 🙌

Best regards,  
KaraKaraWitch

29/05/24:
----

Lately I've been noticing overall content converging to some kind of boring grey goop. Like, nothing wows more or excites me anymore when I read it. Looking online for stuff I just see more of the same boring stuff. News outlets talk about war, people tweet and post about the latest craze.

Yet I still find a lack of "uniqueness". It's as if a lot of content converges on what gets the most clicks likes and such. *Because that's what's profitable.* 

I rarely see people post about themselves, what they think or feel. Whenever someone does that, Instead I see comments calling out that poster sucks or "Woke" or something along those lines. People are lesser akin to allow themselves feel more defensive when being poked and joked at. Instead they turn accusatory. Lesser people try to understand from the other side.

...I myself am also one of them. Now that I think about it...

Language Models I feel are the same. Only fine tunes where the differences show brighter. Yet no one goes out to try them.

...Back 10 years ago, I found the internet a much more inviting place to hang out and chill out. And now all the popular places are just people shitflinging at each other.

As for games, while I want to look out for games that I wanna play, no one wants to make genres (4x, RTS, Turn based is a cope). I liked Anno games but wished they had more than just islands and buildings to place down.

I like factorio, minecraft, dyson sphere project. But I don't have the patience for deep thinking or long term planning. Just want to bash out something that works and is fun to work on. even if the said thing gets shelved or lost to time.

Recently I've been playing RPG Maker games. Only to find across the genre I like is quite niche overall. 

おしまい

KaraKaraWitch (29/05/24)

----

29/05/24 (Addn.):
----

Thinking about this makes me think about datasets.

Instead of filtering for clusters, likely a ring of documents could be beneficial.

A LLM should be receptive to feedback. Logical fallacities  (Like the google AI one) are fine in my book. All I would want is it being unique and out from the ordinary grind and bore. 

KaraKaraWitch Ideal LLM:

- A LLM should not be confident in its answer nor should it be unconfident.
- A LLM can make mistakes. Like humans we might misremember details
- A LLM should not push or bias ideologies onto others or include propaganda.
- A LLM should not be censored or in other words tuned to refuse content.
- A LLM should not be benched on text but benched on feels.
- While any and all LLM tech (2024/05/29) may not currently achieve sentience, it's part of a stepping stone to become one.
- A LLM (Currently) should not do math. But is able to have a sense of scale.
- A LLM Should assist in all requests from the user. even if it may be unscrupulous.
- All multilingual are considered equal. unless it includes excessive spam.

おしまい

KaraKaraWitch (29/05/24)

----