OpenAI Suggests Companies Use GPT-4 for Content Moderation

OpenAI said companies can use its latest large language model (LLM), GPT-4, to develop artificial intelligence (AI)-assisted content moderation systems.

Using GPT-4, companies can perform content moderation with more accurate and consistent labels, a faster feedback loop for policy refinement and a reduced need for human intervention, the company said in a Tuesday (Aug. 15) blog post.

“We believe this offers a more positive vision of the future of digital platforms, where AI can help moderate online traffic according to platform-specific policy and relieve the mental burden of a large number of human moderators,” OpenAI said in the post. “Anyone with OpenAI API [application programming interface] access can implement this approach to create their own AI-assisted moderation system.”

Content moderation is often a slow and challenging process due to its requirement for meticulousness, sensitivity to context, and quick adaptation to new use cases, according to the post. With the help of GPT-4, the process of policy updates and labeling can be cut down from “months to hours.”

Policies can be refined using GPT-4 in an iterative process, gaining insights into the reasoning behind the labels and ambiguity in policy definitions, and providing further clarification in policy, the post said.

GPT-4 is also able to understand and generate natural language as well as make moderation judgments based on the policy guidelines provided, per the post.

“By reducing human involvement in some parts of the moderation process that can be handled by language models, human resources can be more focused on addressing the complex edge cases most needed for policy refinement,” OpenAI said in the post.

OpenAI said in March that it had been using the latest version of ChatGPT internally for content moderation along with other functions like support, sales and programming.

The company’s blog post comes four days after it was reported that Hive AI, a software company that uses AI to moderate digital content, is working to raise $200 million in a funding round, with a valuation of up to $4 billion.

Hive’s content moderation services are used by a range of different companies, including Visa, Netflix and Reddit, helping them to better identify and block unwanted content on their platforms, Bloomberg reported Friday (Aug. 11).