In addition, OpenAI has released two updated GPT-3 base models and plans to retire the original ones by January 2024. The company also announced that fine-tuning support for GPT-4, which can understand both text and images, will be available later this fall. However, no specific timeline was provided for this update.
Key takeaways:
- OpenAI customers can now bring custom data to the lightweight version of GPT-3.5, GPT-3.5 Turbo, improving the AI model’s reliability and specific behaviors.
- Fine-tuning allows companies to make the model follow instructions better, improve the model’s ability to consistently format responses, and hone the model’s output to better fit a brand or voice.
- OpenAI plans to launch a fine-tuning UI in the future with a dashboard for checking the status of ongoing fine-tuning workloads.
- OpenAI has made available two updated GPT-3 base models (babbage-002 and davinci-002), and plans to retire the original GPT-3 base models on January 4, 2024.