OpenAI has introduced upgraded versions of its text-generating AI, GPT-3.5-turbo and GPT-4, featuring a powerful capability called “Function Calling”.
Function Calling enables developers to describe programming functions to the models, which then generate code to execute those functions. These models have been fine-tuned to accurately identify when a function needs to be invoked and provide JSON responses that align with the function’s specifications. Function Calling proves invaluable in developing chatbots that can leverage external tools for answering questions, converting natural language into database queries, and extracting structured data from text.
In addition to Function Calling, OpenAI has also launched an enhanced variant of GPT-3.5-turbo with a significantly expanded context window. The context window refers to the text considered by the model prior to generating additional text. The new GPT-3.5-turbo now offers four times the context length, accommodating up to 16,000 tokens, compared to the vanilla version. Although the price for this extended context window version is double that of the original, OpenAI has reduced the pricing for the standard GPT-3.5-turbo by 25%. Developers can now utilize the model at a cost of $0.0015 per 1,000 input tokens and $0.002 per 1,000 output tokens, equating to approximately 700 pages per dollar.
OpenAI’s ongoing advancements in text generation come at a time of increasing competition within the generative AI landscape. While testing a limited release of GPT-4, which boasts a context window of 32,000 tokens, OpenAI aims to address the limitation faced by models with smaller context windows. Such models tend to forget recent conversation context, leading to off-topic and sometimes problematic responses. The upgraded GPT-3.5-turbo, with its function calling capability, empowers developers to obtain structured data reliably from the model, thereby enhancing its practical applications.
