We realised even with prompting it is very hard to create child appropriate language using AI. Sentences are too long, words too complex, themes sometimes not suited for younger audiences.
This is why we are working on fine tuning open source models for child-appropriateness.
If you want to help or have good datasets to share, check out:
2 comments
[ 4.3 ms ] story [ 14.8 ms ] threadWe realised even with prompting it is very hard to create child appropriate language using AI. Sentences are too long, words too complex, themes sometimes not suited for younger audiences.
This is why we are working on fine tuning open source models for child-appropriateness.
If you want to help or have good datasets to share, check out:
https://oscarstories.com/lora/