OpenAI uses GPT-4 to create an AI-based content moderation system
OpenAI believes that GPT-4 can replace tens of thousands of moderators and still be almost as accurate and more consistent. The developers claim that they are already using the language model to create and refine their own policies, label content, and make decisions.
OpenAI described three main advantages over traditional approaches to content moderation.
First, according to the company, people interpret rules differently, while machines are consistent in their judgments. Manuals can be long and constantly changing. Also, humans need a lot of training to learn and adapt, and language models can instantly implement new rules.
Second, GPT-4 can supposedly help develop a new policy within a few hours. Usually, the process of drafting, labeling, collecting feedback, and finalizing such documents takes weeks or several months.
Thirdly, OpenAI mentions the well-being of employees who are constantly exposed to harmful content, such as videos of child abuse or torture.
