Curated By: Shaurya Sharma
Last Updated: August 24, 2023, 11:07 IST
San Francisco, California, USA
OpenAI, the Microsoft-backed company behind creating ChatGPT, has announced that users of its GPT-3.5 Turbo AI model will now be able to fine-tune and customize the model according to their preferences and specific use cases. This change will allow businesses to effectively deploy these models on a larger scale, and thus align with their specific needs.
Furthermore, the company has revealed that a similar ability will be introduced for GPT-4 later this fall.
As per OpenAI’s early tests, a “fine-tuned” version of GPT-3.5 Turbo holds the potential to match or even outperform “base GPT-4-level capabilities on certain narrow tasks.”And, the data that a certain business uses to fine-tune the model for their own application—will not be shared with OpenAI—ensuring both safety and privacy.
OpenAI has outlined several applications and immediate advantages of fine-tuning the AI model:
By fine-tuning the model, businesses can ensure that it aligns with their particular brand “tone,” making it a more suitable fit for the brand’s identity. “A business with a recognizable brand voice can use fine-tuning for the model to be more consistent with their tone,” OpenAI said.
In simple terms, optimization and fine-tuning enables businesses to shape the model to better fit their operations. For instance, the model can be fine-tuned to respond in German or other languages as required, which may lead to better efficiency.
Consistent Output Formatting
Another immediate benefit that fine-tuning brings is the improvement in the model’s ability to consistently format responses. OpenAI notes that this is a “crucial aspect for applications demanding a specific response format, such as code completion or composing API calls,” and that a developer may use fine-tuning to “more reliably convert user prompts into high-quality JSON snippets that can be used with their own systems.”
Moreover, OpenAI has listed additional benefits, including the ability to handle up to 4k tokens—double that of previous fine-tuned models. Businesses can even reduce the prompt size by up to 90%.