"Safety"? its so restricted that it literally cannot answer basic prompts prompt and refuses to understand basic human input, it just does not listen to the input that is given and does its own thing even with temp 0.1-.05 which is insane.
We need Kids Chats, which bypasses all the adult ignorance and conflicts, connecting cultures, people, and tradition directly among kids all over the universe. This way they bypass any conflict filters adult gain as they grow older, and can with true innocence build strong friendship relations across all of humanity. Disregarding all conflict and impasses which does not exist within those who are unspoiled.
I don't know why people cannot be honest with their product - I worked on with all the models (All the models of the Google are really worst). I check every day whether there is a improvement or not Anthropic n Open AI is doing great, Mistral is good (Meta Llama is good not so good it's in earlier staged) - consistency n quality is important. I don't understand why people invest in useless products n develop it 🙏