OpenAI, the mind behind the controversial artificial intelligence technology ChatGPT, has launched a new feature for the service that would allow users to save preferences and take control of the language model's behavior.
This removes the need for users to write the same instruction prompts to the chatbot every time they interact with the technology. These include inputs such as "Write the answer under 1,000 words" or "Keep the tone of response formal."
ChatGPT's New Custom Instructions
The company behind the chatbot said that the new feature, custom instructions, would let users share anything that they want ChatGPT to consider when generating its responses. They cited an example of a teacher teaching fourth-grade math or a developer specifying the code language they prefer when asking for suggestions.
Additionally, a user can also specify their family size to give ChatGPT the ability to give responses about meals, grocery, and vacation planning by the number given to it, as per TechCrunch.
The new custom instructions feature comes as users can already specify such orders for ChatGPT. However, the new feature is particularly useful if there is a need to input the same prompts when frequently talking with the chatbot.
OpenAI's new feature also works with plug-ins, making it easier to suggest restaurants or flights based on someone's location. The company noted that the latest launch would be available for Plus plan users but not for users in the European Union and the United Kingdom.
Currently, the feature is still in beta feature, and users can test it on the web by clicking on their name, selecting Settings, Beta features, and finally, Opt into Custom instructions. On the other hand, iOS users can access it through Settings, New Features, and Turn on Custom Instructions.
Optimizing the Chatbot's Response Generation
Joanne Jang, who works on model behaviors and products at OpenAI, said that when a user opens ChatGPT, the bot does not know much about them. According to The Verge, she noted that if they started a new thread, the bot would forget everything it talked about in the past.
But the exact way users create their custom instructions and how ChatGPT would interpret them remains a relative mystery. But some have argued that adding these instructions would add to the complexity of a user's query, which could make the chatbot even more prone to being wrong or making up information.
Jang argued that the custom instructions could become more dynamic and interactive with users. When a person tells the chatbot something that enables it to learn about the user, it should be easy to access and be to the user's liking.
The new feature was released just a few weeks after OpenAI released Code Interpreter, which allows users to upload documents, create visualizations based on data provided, and have the chatbot write and run code in Python. These come at a time when there is growing opposition to artificial intelligence and ChatGPT in general, said VentureBeat.
Related Article : James Cameron Says AI 'Weaponization' a Global Risk
