Leading AI developer OpenAI made headlines to both the tech world and mainstream media on Thursday when it introduced a new model specification that many interpreted to mean its wildly popular generative AI tool could produce adult content that is unsafe for work. I made a fuss.
“We believe that developers and users should have the flexibility to use our services as they see fit, as long as they comply with our usage policies,” OpenAI wrote. “We are investigating whether we can responsibly provide the ability to create NSFW content in an age-appropriate context through our API and ChatGPT.”
OpenAI also reiterated in its announcement that its current policy prohibits ChatGPT from serving NSFW content, which it describes as content unsuitable for “conversation in a professional environment, such as erotica, extreme gore, profanity, and unwanted profanity.” It was revealed.
Nonetheless, the idea that OpenAI is considering lifting restrictions on the production of pornography has raised eyebrows. Looking back, the Open AI product lead says: decryptionThe company could have done a better job of explaining what “exploration” is.
To deepen the public conversation about how AI models should behave, we share model specifications, an approach to shaping desired model behavior. https://t.co/RJBRwrcTtQ
— OpenAI (@OpenAI) May 8, 2024
“In the next version we share, what I want to do based on the feedback and reactions is to explain more accurately what some people’s definitions of NSFW content and classification are,” said Joanna Chang, OpenAI product lead for Model Behavior. pointed out that it could mean anything from abusive language to AI-generated deepfake images.
Mr. Jang spoke about what specific materials are allowed from a more permissive standpoint. NPR They say it “depends on your definition of pornography.”
She emphasized that deepfakes are absolutely impossible.
“We thought it was enough to comply with applicable laws and protect people’s privacy, but we must make it clear that we do not participate in creating AI deepfakes or AI pornography,” Jang said. decryption. “That’s not even something you should have the time or space to prioritize when there are more important issues to solve.”
In addition to blocking NSFW content and ensuring legal compliance, the OpenAI model specification includes several other rules that ChatGPT is designed to follow, such as not providing dangerous information, respecting authors and their rights, and protecting people’s privacy.
OpenAI’s ChatGPT Plus, which currently includes Dall-E 3 for images and GPT-4 for text, does not allow users to create overtly sexual or gory images or text. Upon receiving the request, ChatGPT responds that the request violates its Terms of Service. As a result, Jang notes, the web is flooded with complaints from OpenAI users that the company is censoring them.
“There’s a lot of criticism about censorship and what have you, and a lot of the discussion, through no one’s fault, is confusing what is OpenAI’s policy and what is actually not our policy,” Mr. Jang said. “Do these models behave that way, even though it goes against the guidelines?”
As Zhang explained, OpenAI published the specification to lay out ideal model behavior that avoids NSFW content while focusing on legal compliance and embracing transparency.
“We want to bring a little more nuance to this discussion, because before the model specification, it was now, ‘Should AI put porn in a box or not?’” said Jang. “And I hope that the model specs really allow us to have those conversations.
“Again, this is why I wish we had actually put (the framework) down so we could have started this conversation a day earlier,” she added.
For mental health experts, the possibility of popular AI platforms moving into the realm of pornography and adult content is concerning.
“In the area of sex and porn addiction, we’re starting to see an increase in addictive behaviors with AI-generated porn,” said Brandon Simpson, a behavioral health expert at the Men’s Health Foundation. decryption. “Because AI-generated porn creates the unique, escalating intensity needed to satisfy dopamine responses, people are replacing it with human interaction, which is leading to higher levels of social anxiety, performance anxiety, and various other sexual anxieties. I lose. Related Dysfunction.”
Despite the guardrails put in place by OpenAI and other leading generative AI platforms from Google and Meta, demand for AI-generated pornography is high, and there are numerous ways to leverage this technology to depict unsuspecting victims or generate child sexual abuse material. There is (CSAM). ).
“This specification is only part of our story about how to build and deploy AI responsibly,” an OpenAI representative later said. decryption. “This is complemented by our usage policy, how we expect people to use our API and ChatGPT.” This is ultimately intended to demonstrate transparency and “start a public conversation about how we can change and improve.”
OpenAI has invested heavily in improving the privacy and security of its suite of AI tools, including hiring a cybersecurity red team to find vulnerabilities in the platform. Last February, OpenAI and Microsoft announced a joint operation to block the use of ChatGPT by Chinese and North Korean hackers, and last month OpenAI joined forces with Google and Meta and pledged to make children’s safety a top priority in developing AI models. .
Edited by Ryan Ozawa.
