OpenAI is driving the adoption of GPTs, third-party applications powered by its AI models, by enabling ChatGPT users to invoke them in any chat.
Starting today, paid users of ChatGPT, the front end of OpenAI’s chatbot AI, can insert GPTs into the conversation by typing “@” and selecting GPT from the list. The chosen GPT will have an understanding of the full conversation, and can “tag” different GPTs for different use cases and needs – jumping into the conversation with the context of things previously said.
“This allows you to add relevant GPTs with the full context of the conversation,” OpenAI said in a tweet.
The move to make GPTs more discoverable comes weeks after the launch of the GPT Store, a marketplace for GPTs accessible through the ChatGPT dashboard. Creating GPTs does not require programming experience, and GPTs can be as simple or complex as the developer desires. A few of the programs available today include a trail recommender from AllTrails, a code tutor from Khan Academy, and a content designer from Canva.
OpenAI plans to eventually offer monetization to developers who want to sell access to their GPTs. But the company may have to increase traffic first. according to Data From Sameweb, a web analytics company, custom GPT files only account for about 2.7% of ChatGPT web traffic worldwide so far – and custom GPT traffic has been declining month over month since November.
Moderation has proven to be another challenge. In its first week of launch, the GPT store was filled with “romantic” chat apps, some of which were sexually suggestive — a clear violation of OpenAI’s terms. Developers have also been quick to create political campaign bots – such as a chatbot impersonating US presidential candidate Dean Phillips – another clear violation.
OpenAI, which claims to use a combination of human and machine review to determine GPT scores, has since removed some of the offending apps. But if the size of GPTs increases as the company clearly hopes, one imagines the problem will become more severe.