OpenAI is introducing a new tool that will enable corporate clients to personalize GPT-4o, the AI startup’s most potent model, using their own company data.
The change is a result of corporations feeling more pressure to show the return on investment from their AI investments and startups facing increased competition for their commercial AI offerings.
On Tuesday, OpenAI intends to launch the customisation feature, which is commonly referred to as fine-tuning in the AI community. By fine-tuning, current AI models can be trained with more data regarding a certain kind of task or topic. For instance, an AI model may be refined by a skateboard manufacturer to serve as a customer support chatbot that can answer inquiries regarding wheels and board maintenance.
For the OpenAI flagship model, the feature is new; fine-tuning was not previously offered on GPT-4o or its predecessor, GPT-4. Many of the company’s other models, such as GPT-4o small, a more affordable and simplified version of GPT-4o, have been made adjustable by customers.
AI models, both large and tiny, can be customized by a multitude of IT companies. Speaking with Olivier Godement, the head of product for OpenAI’s API, the company is betting that working directly with clients would make it simpler for them to modify its most potent model than utilizing an outside service or a less potent offering.
“We’ve been extremely focused on lowering the bar, the friction, the amount of work it takes to get started,” Godement said.
Customers must upload their data to OpenAI’s servers in order to fine-tune a model. According to John Allard, an OpenAI software engineer who focuses on customization, the training takes about an hour or two on average. According to Allard, users won’t be able to adjust the model at first with graphics or other types of content—only text-based data.
