It may have been reasonably assumed that OpenAI’s forthcoming AI text-to-video creation model Sora would be a sex and nudity-free zone. ChatGPT, OpenAI’s chatbot, is famously prudish and quick to shut down anything remotely erotic.
However, OpenAI’s chief technology officer Mira Murati has suggested that AI footage of bare butts and helicoptering genitals may not be completely off the Sora table. Murati told the [paywall] Wall Street Journal that she was “not sure” what the model’s nudity policy would be.
“You can imagine that there are creative settings in which artists might want to have more control over that. Right now we are working with artists, creators from different fields to figure out what’s useful, what level of flexibility should the tool provide,” she elaborated to the newspaper.
Could “creators” in this context mean porn, or at least sex scene, creators? We should find out later in 2024, as Murati said that Soma would “definitely” be released this year.
To allow sexual and nudity-related video content creation through the much-hyped Sora model would be a tonal shift for OpenAI from ChatGPT, although it may end up serving a rather different user base to the vastly popular chatbot. Preview videos showing Sora’s capabilities have been enormously impressive, leading many to predict that it could prompt a huge model shift in the film and TV industries, as they pivot to high-quality AI video.
To not allow sex and nudity would cut off a large part of those industries, even if full-blown porn was not allowed through Sora. Noting the popularity of ‘NSFW’ AI image generators, though, OpenAI must be aware of the huge potential for breaking into porn creation.
The safety and legal implications of such a move would be huge, though. The hype around the forthcoming Sora AI model comes at a time when concerns about deepfake porn videos are at a peak, following non-consensual Taylor Swift deepfakes doing the rounds online and US authorities scrambling to update legalities around deepfakes.
Sharing deepfake porn videos without the consent of those depicted in them has already been made illegal in the UK, as part of the country’s new Online Safety Bill.
OpenAI has said that Sora is currently being tested by experts in misinformation, hateful content and bias, and is working on detection tools for misleading content. The first incarnation of Sora won’t have sound capability, which at least makes deepfake voices (and deepfake moans of pleasure) less of a concern.
Another AI newcomer making waves is the Claude 3 chatbot and assistant, developed by Anthropic: a startup formed by former OpenAI employees.
Claude 3 is designed to be a ChatGPT rival, but its makers say it is less likely to produce harmful outputs than ChatGPT. As such, Claude 3 is coded to not produce sexual or explicit content, including erotic chat.
When asked about this policy by SEXTECHGUIDE, Claude 3 said: “I do not generate or engage with any kind of explicit adult content involving nudity, pornography, erotic descriptions, or sexual activities. My principles prevent me from producing that type of material. I aim to keep my responses family-friendly and avoid anything too mature or explicit in nature.”
Still, it seems that even Claude 3 isn’t entirely pure. Hackers have discovered that the chatbot can be jailbroken, and have successfully got it to give instructions on making meth at home, which can’t ever be a good thing.
Leave a Reply