OpenAI has updated its API to include function calls, which allows developers to describe functions to GPT-4 and GPT-3.5 and have models create code to execute those functions.
According to OpenAI, function calling facilitates the development of chatbots that can use external tools to convert natural language into database queries and extract structured data from text. The models have been fine-tuned to not only identify instances where functions should be called but also to provide JSON responses that are consistent with the function signatures.
Because function calls are critical, AI models can intelligently connect with external tools and APIs. Developers can access various functions and services by specifying functions for these models. By using external tools to respond to queries, search databases, or extract structured data from unstructured text, this connectivity enables AI models to perform tasks beyond their natural capabilities. Thanks to function calls, AI models become more general and practical, able to tackle complex challenges in the real world.
With the release of gpt-4-0613 and gpt-3.5-turbo-0613, developers can now describe the functions of these models. As a result, models can intelligently generate JSON objects that contain the arguments needed to call these functions. This exciting development provides a more reliable way to connect GPT functionality with external tools and APIs, opening up new possibilities for seamless integration.
These models have developed the ability to identify which function should be activated through careful fine-tuning based on user input. In addition, they have learned to provide JSON answers that match specific function signatures. Developers can now obtain structured data from the model more reliably and consistently using function calls.