On Tuesday, OpenAI announced new features for ChatGPT. This feature allows users to use the @ symbol to draw a custom girlfriend personality called “GPT” into her ChatGPT conversations. This enables a level of quasi-teamwork within ChatGPT between expert roles that was previously unrealistic, bringing collaboration with his team of AI agents within the OpenAI platform one step closer to reality. Masu.
”You can now incorporate GPT into any ChatGPT conversation. Just type @ and select GPT. ” he wrote on OpenAI social media network X. “This will allow us to add the relevant GPT of him, including the full context of the conversation.”
OpenAI introduced GPT in November as a way to create custom personalities or roles for ChatGPT to perform. For example, users can build their own GPTs that focus on specific topics or specific skills. Paid Her ChatGPT subscribers can also freely download Her GPT hosts developed by other Her ChatGPT users through the GPT Store.
Previously, if you wanted to share information between GPT profiles, you had to copy the text, select and paste it into a new chat with GPT, and explain the meaning of the information and the context of what you wanted to do with it. there was. Now, ChatGPT users can stay in the default Her ChatGPT window and bring in Her GPT as needed without losing conversation history.
For example, the “Wellness Guide” GPT, created by experts in human health conditions (of course, this is ChatGPT; if you have a medical problem, always consult a human doctor). We have also created a “Dog Health Guide.” Advisor” accepts questions about dog health.
benji edwards
I started the default ChatGPT chat, pressed the @ symbol, typed the first few letters of “Wellness”, and selected it from the list. That filled in the rest. After asking about food poisoning in humans, I switched to the dog advisor with the @ symbol and asked about dogs.
With this feature, you can also refer to the “Advertising Copywriter” GPT and the “Editor” GPT instead, for example. This means you can have your copywriter write the text, input the rope into your editor GPT, and check it from a different angle. Different system prompts (instructions that define the personality of GPT) can cause significant differences in behavior.
We also attempted to exchange between GPT profiles for creating software and other GPT profiles designed to reference historical technical subjects. Interestingly, ChatGPT does not distinguish between GPTs as different personalities, even if the user changes them. “I’ve done this before” appears even if another girlfriend’s GPT is talking about the output of a previous girlfriend’s GPT in the same conversation history. From that perspective, it’s just ChatGPT, not multiple agents.
From our perspective, this feature appears to be a small step toward a future where GPTs, as independent agents, can work together as a team to perform more complex tasks directed by users. Similar experiments have been done outside of OpenAI (using API access) in the past, but OpenAI has so far resisted ChatGPT’s more agent-like model. As we’ve seen before (first with GPT and now with this), OpenAI seems to be slowly leaning towards that goal itself, but will we see true agent teamwork in delivery services? Only time will tell.
