[gloosx]

We had many examples of AIs which tried to learn from feedback in the public domain. They all quickly becoming racist nazis for some reason.

[AlecSchueler]

What are examples other than Grok which apparently had nazi sympathies hardcoded in the system prompt?