OpenAI Expands Custom Model Program For Enterprise Customers

JAKARTA- OpenAI is expanding its program, Custom Model, to help company customers develop a customized AI-generative model using its technology for certain use, domain, and application cases.

Custom Model was launched last year at OpenAI's inaugural developer conference, DevDay, which offered the company the opportunity to work with a group of OpenAI researchers dedicated to training and optimizing models for certain domains.

Customer "buhans" have been registering in Custom Models since then. But OpenAI says that, by working with this early group of users, they realized the need to expand the program to further "maximize performance." Therefore fine-tuning is helped and customized models.

Fine-tuning is helped, a new component of the Custom Model program, utilizing techniques beyond fine-tuning, such as "additional hyperparameters and various efficient fine-tuning parameter methods on a larger scale," with OpenAI words to allow organizations to prepare pipelines of training data, evaluation systems, and other supporting infrastructure to strengthen model performance on specific tasks.

As for the adjusted models, they are custom models built with OpenAI using basic models and OpenAI tools (miss. GPT-4) for customers who "need to adapt their models more deeply" or "instill new domain-specific knowledge," said OpenAI.

OpenAI provides an example of SK Telecom, a Korean telecommunications giant, who is working with OpenAI to improve its performance in "telecommunications-related conversations" in Korean. Another customer, Harvey who is building legal tools supported by OpenAI Startup Fund, OpenAI's AI-focused arm is working with OpenAI to create a custom model for legal cases that includes hundreds of millions of word legal texts and feedback from licensed expert lawyers.

"We believe that in the future, most organizations will develop personalized customized models for their industry, business, or use cases," OpenAI wrote in a blog post. "With various techniques available to build custom models, organizations of all sizes can develop meaningful and specific personal models to generate more meaningful impacts from their AI implementations."