Leading AI developer OpenAI enraged the tech world and the mainstream press alike on Thursday when it introduced a new model specification that means its wildly popular generative AI tools are unsafe for adult content to work with. It will be allowed to develop.
“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 exploring whether we can provide the ability to responsibly produce NSFW content in age-appropriate contexts through the API and ChatGPT.”
In its announcement, OpenAI also reiterated that its current policy prohibits ChatGPT from serving NSFW content, defined as content that “represents conversation in a professional environment, such as erotica, “Extreme gore, filth, and gross indecency” will not be appropriate.
Nonetheless, the idea that OpenAI was considering lifting restrictions on the creation of pornography raised eyebrows. In retrospect, an OpenAI product leads. Decryptthe company could have better defined what it was doing “food”.
“I think what I'd like to do—based on feedback and reactions, in the next version we'll share—is to be more precise about what some people define as NSFW content and ratings. ” said Joanna Jung, OpenAI Product Lead for Modeling Behavior, noting that NSFW can mean anything from textual profanity to AI-generated deepfake images.
As for what specific content would be allowed under a more legitimate stance, Jung said NPR That it “depends on your definition of obscenity.”
He stressed that deepfakes are completely off the table.
“Following applicable laws and protecting people's privacy, we thought this would cover it — but we must be clear that we are not involved in creating AI deepfakes or AI porn,” Jung said. told. Decrypt. “It's just not something we have the time or the bandwidth to address that should be prioritized.”
In addition to blocking NSFW content and ensuring legal compliance, the OpenAI model specification includes several other rules that ChatGPT is designed to follow, including not providing offensive information, creators and their rights. Respecting, and protecting people's privacy.
Currently, OpenAI's ChatGPT Plus — which includes Dall-E 3 for images and GPT-4 for text — won't allow users to create explicit sexual or gory images or text. When asked to do so, ChatGPT will respond that the request violates its Terms of Service. As a result, Jung pointed out that the web is flooded with complaints from OpenAI users that the company is censoring them.
“There's been a lot of criticism of censorship and what have you, and a lot of that debate is confusing — no one's fault — what OpenAI's policy is and what it actually is,” Jung said. We don't have a policy,” Jung said. “Are these models behaving this way? Even though it's against the instructions?”
As Jang explained, OpenAI publishes speculators to set best practices while embracing transparency, focusing on legal compliance and avoiding NSFW content.
“We want to bring more value to this debate because right now — before the model hypothesis — it was, 'Should AI be creative or not?'” Jung said. “And I'm hoping that through model speculation, we can actually have that conversation.
“Again, this is why I wish I had really laid down [a framework] So we can start this conversation a day earlier.”
For mental health experts, the prospect of popular AI platforms falling into the realm of pornography and adult content is troubling.
“In the field of sex and porn addiction, we're starting to see an increase in addictive behaviors, with AI-driven pornography,” said Brendan Simpson, a behavioral health specialist at the Men's Health Foundation. Decrypt. “Because AI-generated pornography creates the unique and ever-increasing intensity required to elicit a dopamine response, people are diving into it and replacing it with human interactions, which can reduce social anxiety, Performance anxiety, and a variety of sexual related disorders.”
Even with the guardrails put in place by OpenAI and other major generative AI platforms from Google and Meta, AI-generated porn is in great demand, and using the technology to impersonate unsuspecting victims. Numerous methods — or producing child sexual abuse material (CSAM). ).
“The specter of how to build and deploy AI responsibly is only part of our story,” an OpenAI representative later said. Decrypt. “This is complemented by our usage policies, how we expect people to use the API and ChatGPT” — which ultimately aims to demonstrate transparency and “start a public conversation about how to change that.” And can be improved.”
OpenAI has invested heavily in improving the privacy and security of its suite of AI tools, including hiring cybersecurity red teams to find vulnerabilities on its platforms. In February, OpenAI and Microsoft announced a joint operation to stop Chinese and North Korean hackers from using ChatGPT, and OpenAI last month pledged to prioritize child safety in the development of its AI models with Google and Joined Meta.
Edited by Ryan Ozawa.