The purpose of ChatGPT Plugins is to provide specific tools for language models that prioritize safety. These plugins assist ChatGPT in accessing the latest information, performing computations, and utilizing third-party services.

Developers who have been invited off the waitlist can refer to the documentation in order to construct a plugin for ChatGPT. These plugins are then listed in the prompt presented to the language model, along with corresponding instructions on how to utilize each plugin.

Accessing Real-World and Up-to-Date Data

Plugins act as a means for language models to obtain information that is either too recent, too personal, or too specific to be included in the training data. At the user's explicit request, plugins can also enable language models to safely perform restricted actions on their behalf, thereby increasing the overall usefulness of the system.

Plugins offer potential solutions to various challenges encountered with large language models, including issues like generating false information, staying updated with current events, and accessing proprietary information sources with appropriate permissions. By explicitly integrating external data sources such as up-to-date online information, code-based calculations, or information retrieved through custom plugins, language models can enhance their responses by incorporating evidence-based references.

The initial round of plugins has been developed by Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.