The research, dubbed "Deep Ignorance," demonstrates that scrubbing risky content from AI training datasets can create more robust safeguards than traditional post-training safety measures.

@ipontus 1

15 days ago

someeofficial