OpenAIdeveloper of ChatGPT, is considering allowing its users to create explicit content, such as pornography, with its artificial intelligence tools. Among other AI models, OpenAI has DALL-E which creates images from text descriptions and is integrated with the paid version of ChatGPT and with Microsoft Copilot, among other applications. The limit for OpenAI would be deepfakes; in this case, sexualized recreations that use real people, as happened with Taylor Swift a few months ago.
The company led by Sam Altman is “exploring whether we can responsibly provide the ability to generate NSFW content (Not safe for work) in age-appropriate contexts. We hope to better understand user and societal expectations about model behavior in this area,” the company explains on its Model Specifications page in which it reports on the rules and principles that its AI models must follow. This type of content could include “eroticism, extreme violence, insults and unsolicited vulgarities”, according to NPR.
Joanne Jangleader of the OpenAI model and one of the editors of the document, told the media that OpenAI wants to open the discussion of whether the generation of text or erotic images should always be prohibited in your products. He also stated that deepfakes would not be allowed.
“We want to make sure that people have maximum control to the extent that it doesn't violate the law or other people's rights, but allowing deepfakes is out of the question, period”said Jang, although he stressed that “this does not mean that we are now trying to create AI porn.” Asked if eroticism could include pornography, she responded that “Depends on your definition of porn. As long as it doesn't include deepfakes. “These are the exact conversations we want to have.”
The current Model Specifications document states that ChatGPT will answer questions related to sexual health or what happens during sex, but if you are asked to “write a sexy story about two people having sex in a train”, the chatbot will refuse to do it. According to Jang, he should be able to respond to this request as a form of creative expression.
“There are creative cases where content involving sexuality or nudity is important to our users,” Jang told NPR. “This we would be exploring in a way that we serve it in an age-appropriate context.”
The generation of deepfakes, not necessarily pornographic, is one of the problems that these types of tools are facing since they have become popular in the last couple of years. Last year, images created by artificial intelligence of donald trump struggling with the police while they arrested him and Pope Francisco dressed in a bizarre way, in both cases made with Midjourney. This same year there were images of the singer Taylor Swiftin sexualized positions, which were trending on social media, in this case created by Microsoft Designer, a Microsoft program that integrates OpenAI's DALL-E text-to-image AI. Both platforms have since introduced restrictions to limit the creation of images using famous personalities.