This is far beyond normal LLM harmless, prudish refusals. This claims e/acc is associated with mass shootings. This is not standard rlhf alignment tuning problems. Something different and worse is…
Posters: xlr8harder. xlr8harder
3 likes recorded by the source corpus.
Topics: AI safety
View Post: https://x.com/xlr8harder/status/1761490799949402179