OpenAI has announced a new partnership with Scale to support customers fine-tuning it's generative AI models. The partnership aims to empower enterprises to securely customize OpenAI’s advanced models using proprietary data, thus elevating performance, steerability, and application-specific customization.
OpenAI recently introduced fine-tuning capabilities for its GPT-3.5 Turbo model, with plans to extend this feature to the forthcoming GPT-4. Fine-tuning allows businesses to adapt OpenAI's sophisticated models on their own data sets, thus tailoring the technology to unique organizational needs. Importantly, data used in fine-tuning is solely controlled by the customer, ensuring proprietary information remains confidential.
To expand the reach of this fine-tuning utility, OpenAI has now teamed up with Scale AI. The latter is a leader in enterprise-grade AI solutions, with a particular expertise in data enrichment and model evaluation. Now, Scale AI customers can fine-tune OpenAI models via Scale’s enterprise AI platform, benefiting from the company's Data Engine and expertise in customization.
"To get the most value out of our models, companies are looking to use their data to create tailored services and solutions," said Brad Lightcap, COO of OpenAI. "Scale extends our ability to bring the power of fine-tuning to more companies."
Scale has already demonstrated promising results fine-tuning GPT-3.5 for one of its customers, Brex. By leveraging Scale's data annotation and fine-tuning expertise, Brex improved the performance of expense memo generation by 66% compared to the standard GPT-3.5 model.
This boost in accuracy while reducing costs has opened new possibilities for Brex. "Fine-tuning GPT-3.5 has been a game changer for us, enabling us to deliver high-quality AI experiences, comparable to GPT-4, with much lower cost and lower latency," said Henrique Dubugras, CEO of Brex.
Alexandr Wang, Founder and CEO of Scale, reiterated the game-changing potential of fine-tuning for enterprises. "As with software, an incredible amount of value comes from fine-grained optimizations, and fine-tuning is critical for that," he said.
As generative models become increasingly capable, the value of fine-tuning them for specific organizational goals cannot be overstated. With streamlined support from Scale, more companies can now fine-tune advanced models like GPT-3.5 specifically for their use case. This unlocks higher performance, greater efficiency, and capabilities tailored to each business's unique needs. The collaboration primes enterprises to tap into the transformative power of fine-tuned large language models.