this is actually the opposite. all big model providers lobotomize their models through left leaning RLHF