OpenAI, the organization behind the AI powerhouse models like GPT-4 and its predecessor GPT-3.5, is venturing into a strategic partnership with Scale AI, a leader in data labeling. This alliance aims to make it simpler for enterprise developers to fine-tune GPT-3.5 models with custom data.
Fine-tuning allows organizations to adapt AI models for specialized purposes, such as matching brand voice and tone or enabling responses in specific languages. With Scale's Data Engine platform, customers can enhance their data and apply it to fine-tune the GPT-3.5 model to meet their specific needs.
Scale AI confirms that all fine-tuned GPT-3.5 models will undergo a human review to ensure both performance and safety requirements are met, thus providing an additional layer of credibility.
According to Alexandr Wang, CEO of Scale AI, the company’s seven years of expertise in AI and data solutions make it a preferred fine-tuning partner for OpenAI. The partnership aims to deliver high-performing, safe, and impactful AI models to customers.
While OpenAI has hinted at developing its own in-house fine-tuning tool, it’s unclear whether this is intended to supplement or compete with platforms like Scale AI.
OpenAI COO Brad Lightcap views the collaboration as a method to extend the capabilities of OpenAI’s AI models, especially for businesses. "This partnership allows us to extend the benefits of our capabilities to more companies, given Scale's leadership and experience," he stated.
When queried about future fine-tuning possibilities, especially for GPT-4, which is expected to have fine-tuning capabilities in the fall, OpenAI opted not to disclose further details.
With this partnership, OpenAI and Scale AI are setting the stage for enhanced customization and efficacy of AI in the enterprise sector. As OpenAI prepares to offer fine-tuning capabilities for GPT-4, this current collaboration with Scale AI may provide a glimpse into how AI will become more versatile and customizable in the near future.