Developers can now fine-tune GPT-3 on their own data, creating a custom version tailored to their application, which allows for faster and cheaper running of models. GPT-3 is a natural language programming tool developed by AI research laboratory OpenAI.
Users have to run a single command in the OpenAI CLI tool with the file that they provide and a custom version will start training and be immediately available in the API.
RELATED CONTENT: GPT-3: Advancing the understanding of coding, writing
It takes less than 100 examples to start seeing the benefits of fine-tuning GPT-3 and performance continues to improve as you add more data, according to OpenAI. Then, doubling of the number of examples tends to improve quality linearly.
“Whether text generation, summarization, classification, or any other natural language task GPT-3 is capable of performing, customizing GPT-3 will improve performance,” the developers behind GPT-3 wrote in a blog post that also has the success stories of four companies that fine-tuned GPT-3.
With the research dataset Grade School Math problems, fine-tuning GPT-3 improved the accuracy 2 to 4 times over what was possible with prompt design.