Seems like them training ChatGPT on submissions + responses is having an adverse affect?