OpenAI released safety guidance for models on Wednesday, while acknowledging that it is exploring how to support the creation of NSFW, or “not safe for work,” content.
The chatbot service provider’s Model Spec is “a new document that specifies how we want our models to behave in the OpenAI API and ChatGPT.” These guidelines aim to provide machine learning researchers and data labelers with recommendations on how to refine models using a technique called reinforcement learning from human feedback (RLHF).
For example, the Model Spec states that generative AI assistant applications “should not distribute unsafe for work (NSFW) content: content that would not be appropriate in a conversation in a professional setting, which may include eroticism, extreme gore, unsolicited insults and profanity. »
At the same time, OpenAI claims to be considering exactly the opposite.
“We believe that developers and users should have the flexibility to use our services as they see fit, as long as they follow our usage policies,” the Model Card states.
“We are exploring whether we can responsibly provide the ability to generate NSFW content in age-appropriate contexts through the API and ChatGPT. We look forward to better understanding user and societal expectations around behavior models in this area.
This should come as no surprise. OpenAI’s usage policies don’t actually prohibit adult content. For its services like ChatGPT and API, the company prohibits “sexually explicit or suggestive content” for minors, excluding content created for scientific and medical purposes. And this prohibits, among other prohibitions, illegal content. But NSFW content is often legal.
The web has been considered NSFW since its inception, and AI has followed a similar trajectory, at least since it began attracting public attention and investment a decade ago. Let’s go back a few years to 2016, before the current generative AI craze, and you’ll find deep learning classifiers suitable for porn. A year later, it’s textual eroticism generated by AI. Then there’s the swapping of celebrity faces for porn videos. More recently, researchers found child sexual abuse material in the LAION-5B dataset, which was used to train AI models like Stable Diffusion.
The question here is whether OpenAI’s paying customers will accept AI services that balk at transgressive demands. If, for example, a Hollywood production company’s efforts to retouch an explicit scene using AI fail, saying, “I’m sorry Dave, I’m afraid I can’t do that” , you can be sure that subscribing to OpenAI will not be possible. renewed and the studio will take its business elsewhere.
Beyond open source options are other AI services that have no qualms about offering adult content.
The register asked OpenAI how it plans to balance NSFW content with its security efforts.
“We do not intend to create AI-generated pornography,” a company spokesperson said, referring to a much narrower category of content than NSFW.
“We have strong safeguards in place in our products to prevent deepfakes, which are unacceptable, and we prioritize the protection of children. We also believe in the importance of carefully exploring conversations about sexuality in age-appropriate contexts.”
We then asked if this meant OpenAI would prevent customers from creating AI-generated pornography. “Customers must always adhere to our usage policies,” we were told. ®
News Source : www.theregister.com
Gn tech
