One thing I am exploring is that, with LLM (and enough $ to pay the tokens) you can filter out offenses and hate speech in player communications.
All typed message is effectively translated by the LLM as “coach speech” and what is put in the UI is the LLM output. Sure, it is not exactly the player words and can get expensive, but I think it can be a good solution
One thing I am exploring is that, with LLM (and enough $ to pay the tokens) you can filter out offenses and hate speech in player communications.
All typed message is effectively translated by the LLM as “coach speech” and what is put in the UI is the LLM output. Sure, it is not exactly the player words and can get expensive, but I think it can be a good solution