As more companies compete in the field of AI that generates text, OpenAI is improving its models that create text and making them more affordable.
OpenAI has announced the launch of new versions of its text-generating AI called GPT-3.5-turbo and GPT-4. GPT-4 is the latest version and it has a special feature called function calling. This means that developers can tell GPT-3.5-turbo and GPT-4 what they want to do with programming and the models will create the code to make it happen.
For example, with function calling, developers can make chatbots that answer questions by using other tools, change normal language into commands for a database, and extract important information from text. OpenAI explains that these models are designed to know when a function needs to be used and they can give back organized data in a special format called JSON.
Apart from function calling, OpenAI is also introducing a version of GPT-3.5-turbo that can consider a lot more text before generating its own. The amount of text it considers is called the context window, and models with small context windows tend to forget what was said before and might go off-topic or say things that cause problems.
In simpler terms, OpenAI plans to make small improvements to their current models instead of creating completely new models. After releasing GPT-4 in March, OpenAI’s CEO, Sam Altman, mentioned at a conference that they haven’t started working on the next model yet. He explained that they have many tasks to complete before they begin developing the new model.
