Very cool new #LLM study in @natureportfolio.bsky.social 📢

LLMs, like humans, exhibit in-group favouritism & out-group degradation.

Both in the lab and in real human-LLM chats.

We *must* carefully curate training data to mitigate this bias 👩‍💻

https://www.nature.com/articles/s43588-024-00741-1

Comments