#Create community dataset to finetune models
3 messages · Page 1 of 1 (latest)
Absolutely, as long as it's completely transparent and parsed for PII and such for privacy concerns.
On the one hand, this would be a cool data source and might be able to help our own future plans and maybe even other projects in the future, too.
On the other hand, we already get a number of GitHub issues and messages in Discord about privacy concerns with the extremely limited stuff the codebase is already doing.
Something like this would need to be behind some serious opt-in barriers and documented heavily.