Finetuning of GPT-3 Is Now Officially Allowed, says OpenAI

OpenAI has announced that its customers now have the opportunity to create custom versions of GPT-3, a model that can generate human-like text and code. The new GPT-3 fine-tuning capability enables customers to train GPT-3, to identify definitive patterns for tasks and workloads like content generation, classification, and text summarization in particular domains.

The fine-tuned models demonstrate 2x accuracy when tested on questions from the same dataset. In addition to that, fine-tuning can be advantageous, as long as it can enable companies to keep custom GPT-3 models “fresher.”

Previously, OpenAI partnered with Microsoft to introduce the Azure OpenAI Service, an offering designed to give enterprises access to GPT-3 and its derivatives along with security, compliance, governance, and other business-focused features.