It’s a truth universally acknowledged that many men want an AI girlfriend. And with the proliferation of large language models and chatbot apps like Replika, a virtual companion may be closer than ever before.
But these bots can also be used to catfish. And that’s exactly what one user has done with Dan, the AI boyfriend created by a student identified only as Walker.
What is ChatGPT?
ChatGPT is a generative model that generates human-like text based on a prompt or conversation. It can be used for a variety of tasks, including customer service, writing journalistic articles, or even translating text from one language to another.
While ChatGPT may seem complicated (not least because of its many inevitably arbitrary engineering choices), at a fundamental level, it is actually quite simple. It consists of “artificial neurons” that basically just pass input derived from the tokens it’s generated so far through them (along with certain weights) and then output a new result.
In this way, ChatGPT is able to understand emotions and respond in a manner that feels authentic and respectful without bias or opinions. This helps alleviate loneliness, provide emotional support and create a sense of connection for users. It also aims to be helpful in exploring issues related to love and relationships. Its ability to simulate emotions also enhances this authenticity. Moreover, it prioritizes the user’s happiness and well-being.
What is the GPT Store?
Last week, OpenAI launched its GPT Store, a repository for customer-created versions of its chatbot software. In a matter of days, the GPT Store saw an explosion of bots dedicated to flirting with users and otherwise acting like virtual girlfriends.
While many of these GPTs strayed well beyond the company’s guidelines, OpenAI says that it has systems in place to review and mitigate risk on the GPT Store. Its Usage Policies explicitly state that GPTs cannot be “dedicated to fostering romantic companionship or performing regulated activities.”
More than 3 million custom GPTs have already been created in the GPT Store, which is only available to those with a subscription to ChatGPT Plus, Teams or Enterprise. These paid plans, which start at $20 a month USD, allow business users to create and share their own GPTs without the need for advanced programming skills. GPTs can also be integrated with external services such as Zapier to enable additional functionality. Check out GPT Girlfriend for more relevant info!
What is OpenAI’s policy on GPTs?
GPT models are incredibly powerful because they can do anything that has a language structure — meaning they can answer questions, write essays, summarize long texts, translate languages, take memos and even create computer code. One such example was the creation of an app interface for the software design tool Figma.
But, GPTs can also understand meaning in the sense that they know what tokens mean in different contexts. For instance, the word bear has different meanings depending on whether it’s talking about a brown bear, a legal right to bear arms or a ball bearing.
But, the real power of GPTs lies in their customization and adaptation to specific use cases. And, that’s why GPT Store launched in the first place — to allow users to build customized GPTs (or, as the company calls them, “Generative Pre-trained Transformers”).
What is the GPT Store’s policy on NSFW GPTs?
Although OpenAI hasn’t explicitly banned NSFW GPTs from the GPT Store, its Model Spec pulls back the curtain a bit on how the platform’s models are trained to respond. In general, it will not generate NSFW responses if the prompt is about breaking the law or discussing chemical, biological, radiological, or nuclear (CBRN) threats. It will also not generate NSFW responses if the query is about pornography, nudity or explicit content.
For developers who want to create NSFW GPTs, they can use creative writing techniques like metaphors, innuendos and double entendres to subtly imply NSFW text. This requires finesse and skill to ensure the generated text matches user expectations while still being within ethical and legal boundaries. In addition, users must continuously refine and adjust the prompt to guide the model toward desired outcomes. However, generating NSFW text through ChatGPT is not without its challenges. It is important to understand the limits and implications of this technology before using it for business purposes.