OpenAIdisplayA test content review function has been added to its large-scale natural language model GPT-4, allowing users to build their own content review system, which will help improve the overall content accuracy.
This feature will allow users to build their own content review functions, allowing various artificial intelligence application services built through GPT-4 to be used for content filtering and inspection, thereby reducing the workload that originally required a lot of manpower.
By designing certain content review rules, OpenAI said it will be able to enable GPT-4 to review whether the content meets the review requirements or violates the terms of service. This will make the review method that originally required content filtering systems and a large amount of human intervention simpler.
The content review system can be adjusted at any time according to changes in review regulations. It can also speed up the time required for content review and further reduce the proportion of human intervention, allowing the service to operate more efficiently and automatically.
However, from the perspective of the overall operational process, even if the system can automatically review content, misjudgments may still occur. Therefore, OpenAI emphasizes that this function is not intended to completely replace traditional manpower, but to make human resource arrangements more efficient.
According to OpenAI's assessment, this feature can reduce the time it takes to filter online content, which originally took six months to complete, to about one day. This is especially true given the large amount of content generated daily on online platforms. Using artificial intelligence to assist with content review will bring greater advantages, while also reducing the potential impact of traditional manual review of large amounts of content, such as creating a mental burden on those who face reviewing large amounts of content daily, or leading to inertia leading to a lack of comfort in content review.
OpenAI said that by using artificial intelligence to assist in content review, traditional manual review work can focus on content that needs further judgment to determine whether it complies with regulations, rather than spending a lot of physical effort to handle a large number of trivial and intuitive content checks.
