Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

And it begins. OpenAI mulls NSFW AI model output

May 10, 2024 - theregister.com
OpenAI has released guidelines on model safety, outlining how it wants its models to behave in the OpenAI API and ChatGPT. The guidelines, intended for machine learning researchers and data labelers, recommend how to fine-tune models using reinforcement learning from human feedback (RLHF). The Model Spec states that AI assistant applications should not serve NSFW (Not Safe For Work) content, which includes erotica, extreme gore, slurs, and unsolicited profanity. However, OpenAI is also considering providing the ability to generate NSFW content in age-appropriate contexts, as long as it complies with usage policies.

Despite OpenAI's usage policies not explicitly disallowing adult content, the company has stated it has no intention to create AI-generated pornography and has strong safeguards to prevent deepfakes. The company is prioritizing the protection of children and exploring conversations about sexuality in age-appropriate contexts. Customers are still required to adhere to OpenAI's usage policies.

Key takeaways:

  • OpenAI has released model safety guidance, which includes recommendations for fine-tuning models using reinforcement learning from human feedback (RLHF) and guidelines for content generation, particularly NSFW content.
  • While the Model Spec advises against generating NSFW content, OpenAI is considering the possibility of responsibly providing the ability to generate such content in age-appropriate contexts, as long as it complies with their usage policies.
  • OpenAI's usage policies do not disallow adult content, but they do prohibit sexually explicit or suggestive content for minors and unlawful material. However, the company is clear that it has no intention to create AI-generated pornography.
  • OpenAI is aware of the potential business implications of restricting certain types of content, and is considering how to balance user flexibility with safety and ethical considerations.
View Full Article

Comments (0)

Be the first to comment!