Azure OpenAI Content filtering results difference between English and other Language

Question

Hi, I know it is a bit vague question, but is there any content filtering result difference between English and other language such as Chinese and Japanese? If it is English, the results seems reasonable, but for other languages, we feel the filtering results are not always accurate. I would appreciate it if someone could share any insight on the point.

Answer 1

@KT Thanks for the feedback, could you please share some examples of the bad result from your end so that we can take a look at it? We will investigate it and see if we can improve the result.

As an AI model developed by OpenAI, GPT-4's content filtering capabilities can vary between languages due to the training data it was fed and the model's proficiency in understanding the context of different languages.

English is the primary language for most AI models, including GPT-4, because the majority of the internet's content is in English. When it comes to other languages like Chinese or Japanese, the model might not perform as efficiently. This is because the amount of non-English training data is usually less, and the model may not understand the nuances, context, and grammar rules of these languages as well as it understands English. This can result in less accurate content filtering results.

We are looking forward to your response and we can see the next steps we can do. I hope this helps.

Regards,

Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.