OpenAI’s enterprise customers who want to use generative AI but focus on accuracy may be able to achieve this goal through new customization options in the GPT-4 API.
They include the ability to connect to third-party platforms to share fine-tuning information, the ability to save fine-tuned models during additional training periods without retraining the entire model, and a new user interface for comparing model performance and quality. Launched in 2020, the OpenAI API allows the company to connect and build on top of its large language models, followed by fine-tuning GPT-3.5 last August and eventually GPT-4 as well.
Highly customized models can deliver better results in-house. They can also remain responsive to the public and avoid embarrassing situations, such as a ChatGPT-powered Chevy dealership bot saying it would sell a car for $1.
Today, OpenAI COO Brad Lightcap said in a speech Bloomberg According to the interview, 2024 is the “year of the enterprise” as companies start to launch more use cases using the basic model, and OpenAI is seeing “huge growth” in this area. “We increasingly feel that the market is pulling us toward real applications that deliver real business outcomes through a broad focus on AI enablement, and we are ready to support our customers in doing so,” Lightcap said.
Lightcap has previously warned executives not to expect AI to solve all their problems, warning that generative AI is still in an experimental stage.
OpenAI also officially announced assisted fine-tuning, where OpenAI employees help customers fine-tune GPT-4 as part of its custom model program. The Custom Models program allows selected organizations to work with OpenAI researchers to train custom GPT-4 models based on their specific needs.
Fine-tuning allows pre-trained models like GPT-4 to learn more from smaller, more targeted datasets, such as a company’s investment portfolio. The first third-party integration is with the Weights and Bias platform.
While OpenAI is best known for ChatGPT, which appeals to everyday consumers and brands who want to create their own versions of chatbots, the company has partnered with startups and more established organizations to bring AI-driven applications and services to the market. Enter the mainstream.
Some of the earliest darlings of the generative AI era, such as copywriting platform Jasper and video creation company Descript, have built their services using OpenAI’s models.