OpenAI and Scale AI announce collaboration to offer fine-tuning services for GPT-3.5 to empower enterprises in customizing models.

2023-08-25

OpenAI announced that it will provide users with new built-in support for the GPT-3.5 Turbo large language model (LLM), enabling businesses to train models and run them at scale using their proprietary data. Subsequently, the company named Scale AI as its "preferred partner" to offer fine-tuning services for GPT-3.5. Alexandr Wang, the founder and CEO of Scale AI, stated in a press release, "We are excited to collaborate with OpenAI to empower model performance and help every business effectively utilize artificial intelligence according to their unique needs. Simply using the best LLM, such as GPT-3.5, for prompts is not enough to achieve accurate and efficient customized models. Just like software, subtle optimizations bring significant value, and fine-tuning plays a crucial role." Scale AI provides expertise in fine-tuning. According to their press release, this collaboration "combines OpenAI's advanced base model GPT-3.5 with Scale's expertise in fine-tuning and industry-leading Data Engine to help each company create custom state-of-the-art models based on their specific business needs." Scale's Data Engine accelerates model development by generating prompts and ranking model outputs. Scale stated that they have already performed fine-tuning on numerous commercial and open-source models. "As the preferred fine-tuning partner for GPT-3.5, we are delighted to leverage their powerful API to help more businesses build powerful custom LLMs, improving efficiency while reducing costs," the press release stated. Brad Lightcap, the Chief Operating Officer of OpenAI, said, "Scale expands our ability to bring fine-tuning capabilities to more companies. With their experience in enterprise AI, they help businesses better apply OpenAI models to meet their unique needs." Scale AI is a well-known unicorn company and a data annotation service provider. It achieved a valuation of $7 billion in April 2021. Despite a 20% reduction in its 700-person team in January, it made a comeback in May with the launch of a full suite of generative AI platforms. Now, it cites a fine-tuning case study with fintech company Brex, which used LLM to generate high-quality expense memos, helping alleviate the burden of compliance for employees. According to Scale's press release, the Brex team had been using GPT-4 to generate memos but wanted to explore if they could improve cost and latency while maintaining quality by using the fine-tuned GPT-3.5 model. The press release mentioned, "By using the GPT-3.5 fine-tuning API and combining it with Scale's Data Engine annotated with Brex data, we found that the fine-tuned GPT-3.5 model outperformed the original GPT-3.5 Turbo model in 66% of cases." Henrique Dubugras, the CEO of Brex, said, "Our ongoing collaboration with OpenAI and Scale AI puts us at the forefront, enabling us to enhance employee compliance and help finance teams complete their books faster by applying state-of-the-art technology. Fine-tuning GPT-3.5, in particular, has been a game-changer for us, allowing us to provide a high-quality AI experience comparable to GPT-4 at lower costs and latency. It has unlocked a range of new capabilities that were previously unattainable for us."

RECENT AI NEWS

RECENT AI TOOLS