LocalLLaMA

3011 readers

50 users here now

Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.

Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.

As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.

founded 2 years ago

MODERATORS

SkySyrup@sh.itjust.works

pax@sh.itjust.works

noneabove1182@sh.itjust.works

Smokeydope@lemmy.world

MonsterBug@sh.itjust.works

I don't want them to, but Snapchat would have a great dataset to make an actually funny LLM. (sh.itjust.works)

submitted 1 week ago by pebbles@sh.itjust.works to c/localllama@sh.itjust.works

4 comments fedilink hide all child comments

My groupchats use those react emoji all the time. Maybe they could train a model to classify with those. Then use that classifier to help RL models into being funny.

All my funniest groupchats are on Snapchat.

I don't think this would be ethical, but it could be effective.

top 4 comments

sorted by: hot top controversial new old

[–] General_Effort@lemmy.world 1 points 6 days ago* (last edited 6 days ago) (1 children)

What wouldn't be ethical about it? Seems implausible that any private details would be leaked.

ETA: Sounds like a clever idea.

[–] pebbles@sh.itjust.works 1 points 6 days ago

Well they get to use my friends social inertia against me. Hard to move off of snap even if I wanted to and so they can just change the TOS and get my data.

But they don't have to do it that way.

It could be a pretty bomb model though.

[–] Sims@lemmy.ml 2 points 1 week ago (1 children)

Humor is rather subjective, but if you feel you have THE material for "Humor bot 2000", then there are ways of fine-tuning your own model with exactly that kind of Humor.

I haven't tried them, but if they suck ass to setup, maybe your snapchat buds could help out.

https://github.com/xming521/WeClone (can be used on free Colab apparently) https://github.com/mindverse/Second-Me

These are meant to create an 'Avatar' to represent a user, but you can train them with anything. Not sure how they handle audio/pics or the important Timing parts involved in humor, but both finetune a model with your personal training set, so extract your funny shit in the right format and go for it ;-)

[–] pebbles@sh.itjust.works 1 points 1 week ago

Yo thanks for the resources! I never even considered doing it myself. Kinda scared it would reflect me too well yk lol.

I may give this a shot.