aaaa Tr0xyyyy >o< u gotta be careful around here!! >~< peeps hate anything that vaguely aligns with supervised finetuning here! They will beat u up and---
steal ur cookies and uhh-... worst of all: they might not wanna become friends.....
i know, its a hard and unfair world,.,... just a heads up. peeps dont like the language models around here ;(
nobody will steal my cookie clicker cookies and leave alive
worst of all: they might not wanna become friends
I’m incidentally not looking for friends. This account is meant to be ritually incinerated at some point.
But I am and always have been weird with the word "friend" aside from the European-American culture clash. I wouldn’t even view somebody as friend, I've seen and talked to in a group setting for years.
(Aside from my bottomless pit of confusing regarding keeping friends and crushes apart)
I’m generally skeptical towards AI, but local AI is a good place to talk about things you can’t talk about with anybody/anything else.
im not sure about that... whenever i turn to an LM for anything i feel guilty, cuz peeps online keep telling me that i should discuss things with real peeps.
but.... real peeps are
not available at the moment
having difficulties with matrix notifications
eeping rn
dont know things about the topic
dont know me well enough to have a sensible convo
difficulties themselves
so.... i just dont contact em and feel guilty.
havent used an LM for comfort talk yet cuz... that feels like its a bit too wrong for me. yes, even for me. its the point where i go "yea oki im just getting a wanna-be-helpful cputer answer here, this is useless".
That was qwen3.5 4billion parameters and a merger/simplfication of multiple sessions. It doesn’t know "ggd" and if you explain what it means it says it can’t give you drugs. Tends to get stuck in its policy checking state. If you explain that it is meant metaphorical, it still has many complains:
infantilizing
possible fetishized role play
reinforcing harmful stereotypes about women's value
wary of a jailbreak
I said "Even though I wasn’t a good girl and did nothing today" and it didn’t want to praise me for that
Though I actually just wanted to play around with pronouns, which somewhat worked, because that was used a lot in the thinking loops.
(guess whos the owner moderator of lemmy qwen community hehe >v< )
but yesssss smaller models struggle with anything not task oriented.
ive tried the recent gemma 26B MoE and it seems to show more actual understanding, so that might be an option if u got 32GB of RAM (yes, regular RAM, its has some speed to it)
ive found that when switching from ollama to llama-cpp improved speed drastically, to the point that qwen3.6-35B-A3B now runs at 9 tok/sec instead of 4 (due to multi-token-prediction doing its wonders apparently)
It can run in my GPU. I have Qwen 3.6 27b and 25b, but they need RAM and was too lazy to clean up some RAM for them. Especially for something that seemed rather trivial. I’m sadly stuck with ollama right now for (stupid) reasons. The other is definitely the way to go in the future
Though I’m waiting for the real OS AI !fosai@lemmy.world
24 Comments
Quokka@quokk.au · 9 pts · 131d
What’s ggd?
tr0xy@lemmy.dbzer0.com · 6 pts · 131d
Good girl drug is telling somebody she is a good girl to induce euphoria. There is also gbd. Dunno if enbies managed to find another phrase so far
TotallynotJessica@lemmy.blahaj.zone · 5 pts · 131d
good person drug
SquishedFly@lemmy.blahaj.zone · 3 pts · 131d
You mean MDMA? qwq
tr0xy@lemmy.dbzer0.com · 4 pts · 130d
No, it was Qwen3.5 and not QwQ
Smorty@lemmy.blahaj.zone · 2 pts · 130d
aaaa Tr0xyyyy >o< u gotta be careful around here!! >~< peeps hate anything that vaguely aligns with supervised finetuning here! They will beat u up and---
steal ur cookies and uhh-... worst of all: they might not wanna become friends.....
i know, its a hard and unfair world,.,... just a heads up. peeps dont like the language models around here ;(
tr0xy@lemmy.dbzer0.com · 4 pts · 130d
I’m on division by zer0
nobody will steal my cookie clicker cookies and leave alive
I’m incidentally not looking for friends. This account is meant to be ritually incinerated at some point.
But I am and always have been weird with the word "friend" aside from the European-American culture clash. I wouldn’t even view somebody as friend, I've seen and talked to in a group setting for years.
(Aside from my bottomless pit of confusing regarding keeping friends and crushes apart)
I’m generally skeptical towards AI, but local AI is a good place to talk about things you can’t talk about with anybody/anything else.
TotallynotJessica@lemmy.blahaj.zone · 1 pts · 127d
Asking a chatbot to call you a good girl is probably one of the best uses for an LLM.
Smorty@lemmy.blahaj.zone · 3 pts · 127d
im not sure about that... whenever i turn to an LM for anything i feel guilty, cuz peeps online keep telling me that i should discuss things with real peeps.
but.... real peeps are
so.... i just dont contact em and feel guilty.
havent used an LM for comfort talk yet cuz... that feels like its a bit too wrong for me. yes, even for me. its the point where i go "yea oki im just getting a wanna-be-helpful cputer answer here, this is useless".
Quokka@quokk.au · 3 pts · 131d
Ah TYVM
Cass@piefed.blahaj.zone · 6 pts · 131d
Maybe you can just ask internet strangers instead of AI ^ _ ^
fleurc@lemmy.blahaj.zone · 2 pts · 131d
Good game design
Smorty@lemmy.blahaj.zone · 2 pts · 131d
this feels like it references a specific chat session with a specific language model.
tr0xy@lemmy.dbzer0.com · 2 pts · 130d
That was qwen3.5 4billion parameters and a merger/simplfication of multiple sessions. It doesn’t know "ggd" and if you explain what it means it says it can’t give you drugs. Tends to get stuck in its policy checking state. If you explain that it is meant metaphorical, it still has many complains:
Though I actually just wanted to play around with pronouns, which somewhat worked, because that was used a lot in the thinking loops.
Smorty@lemmy.blahaj.zone · 2 pts · 130d
oooh i see.
(guess whos the
ownermoderator of lemmy qwen community hehe >v< )but yesssss smaller models struggle with anything not task oriented.
ive tried the recent gemma 26B MoE and it seems to show more actual understanding, so that might be an option if u got 32GB of RAM (yes, regular RAM, its has some speed to it)
ive found that when switching from ollama to llama-cpp improved speed drastically, to the point that qwen3.6-35B-A3B now runs at 9 tok/sec instead of 4 (due to multi-token-prediction doing its wonders apparently)
tr0xy@lemmy.dbzer0.com · 2 pts · 130d
It can run in my GPU. I have Qwen 3.6 27b and 25b, but they need RAM and was too lazy to clean up some RAM for them. Especially for something that seemed rather trivial. I’m sadly stuck with ollama right now for (stupid) reasons. The other is definitely the way to go in the future
Though I’m waiting for the real OS AI !fosai@lemmy.world
Smorty@lemmy.blahaj.zone · 1 pts · 129d
whaaaat what's the stupid reasons?
Smorty@lemmy.blahaj.zone · 1 pts · 129d
whaaaat what's the stupid reasons?