OpenAI is fine-tuning its Custom Model, to help enterprise customers develop tailored generative AI models using its technology for specific use cases, domains, and applications.
At OpenAI’s first developer conference, DevDay, the company unveiled its Custom Model program, allowing businesses to collaborate with a team of dedicated researchers to refine and optimize AI models for specific purposes. Since its launch last year, the program has attracted “dozens” of customers.
However, OpenAI recognized the necessity of expanding the initiative to enhance performance further. Consequently, the company is introducing assisted fine-tuning and custom-trained models to meet this demand.
According to OpenAI, the assisted fine-tuning component of the Custom Model program techniques such as “additional hyperparameters and various parameter efficient fine-tuning methods at a larger scale.”
This enables organizations to set up data training pipelines, evaluation systems, and other supporting infrastructure toward bolstering model performance on particular tasks.
As for custom-trained models, they’re custom models built with OpenAI — using OpenAI’s base models and tools (e.g. GPT-4) — for customers that “need to more deeply fine-tune their models” or “imbue new, domain-specific knowledge,” said OpenAI.
OpenAI provided examples of how companies utilized the Custom Model program to enhance their AI capabilities.
One instance involved SK Telecom, a prominent Korean telecommunications company, collaborating with OpenAI to fine-tune GPT-4 for better performance in Korean telecommunications-related conversations.
Another case involved Harvey, a company developing AI-driven legal tools with backing from the OpenAI Startup Fund. The company partnered with OpenAI to create a custom model for case law. This model incorporated extensive legal text and feedback from licensed expert attorneys, totaling hundreds of millions of words.
OpenAI wrote in a blog post, “We believe that in the future, the vast majority of organizations will develop customized models that are personalized to their industry, business, or use case.”
It added, “With a variety of techniques available to build a custom model, organizations of all sizes can develop personalized models to realize more meaningful, specific impact from their AI implementations.”
OpenAI’s Growth And Pressure To Maintain The Momentum
OpenAI is experiencing significant growth, with reports indicating that it is approaching an impressive annualized revenue of nearly $2 billion. However, there’s undoubtedly internal pressure to sustain this momentum, especially as the company collaborates on a reported $100 billion data center project with Microsoft.
Given the persistent high costs associated with training and deploying flagship generative AI models, revenue growth is crucial. Consulting services, such as custom model training, may serve as a vital strategy for maintaining revenue while OpenAI strategizes its next steps.
Fine-tuned and custom models offer the potential to alleviate strain on OpenAI’s model serving infrastructure. These tailored models are often smaller and more efficient compared to general-purpose models.
As the demand for generative AI continues to rise, such customized solutions undoubtedly emerge as an appealing option for OpenAI, which has historically faced challenges related to computing capacity.
In addition to announcing the expansion of the Custom Model program and custom model building services, OpenAI introduced new features for developers working with GPT-3.5.
These features include a revamped dashboard for comparing model quality and performance, support for integrations with third-party platforms (beginning with the AI developer platform Weights & Biases), and improvements to tooling.
Details about fine-tuning for GPT-4, which was launched in early access during DevDay, remain undisclosed for now.
OpenAI is revolutionizing AI with its Custom Model program, addressing diverse industry needs. As demand surges, the company’s strategic expansions and innovations promise even greater advancements in tailored AI solutions.
READ ALSO: Trump’s Bid To Dismiss Georgia Election Subversion Case Rejected
