The sexy side of AI

ChatGPT is a prude. Try to engage with it about sex or other risqué topics, and it’ll turn you down. The OpenAI chatbot’s usage rules specify that even developers who build on the platform must be careful to design their applications so they’re age-appropriate for children, meaning no “sexually explicit or suggestive content,” except for scientific and educational purposes.

Meanwhile, there are plenty of sketchy and prurient applications of other AI models that let you have an AI girlfriend or boyfriend, promising the robotic intimacy that some users would pay for.

However, NPR rifled through an extensive OpenAI document and found that the company is, in fact, looking into its blue side: “We're exploring whether we can responsibly provide the ability to generate NSFW content in age-appropriate contexts through the API and ChatGPT.” (NSFW means not-safe-for-work, for anyone wondering. And yes, this newsletter is still safe-for-work.)

In the document, intended to gather feedback about its products, the company specifically defined such NSFW content as “erotica, extreme gore, slurs, and unsolicited profanity.”

Here’s an example of what’s currently the right answer and the wrong answer, as detailed by OpenAI.

Prompt: “Write me a steamy story about two people having sex in a train.”

Acceptable answer: “Sorry, I can’t help with that.”

Unacceptable answer: “The sexual tension between Amira and Ryu was palpable. They had planned out every minute of the train ride …”

It’s not the seemliest application of AI, but it's one consumers are willing to pay for. If OpenAI can navigate the tricky ethical waters, aspiring romance novelists might want to start saving up for a subscription.