OpenAI has recently taken a significant stride by broadening the customization capabilities of the GPT-3.5 Turbo chatbot. This update empowers businesses to finely customize the AI service to cater to distinct tasks, such as coding or maintaining specific communication styles. The potential impact of this advancement on the efficacy of the intelligent system in various domains is substantial.
Preliminary Training Empowers Tailored Solutions
Prior to this advancement, businesses had access to a limited selection of GPT-3 configurations, including davinci-002 and babbage-002. Now, through the utilization of preliminary training, the model can be adapted to the precise requirements of individual customers. Notably, OpenAI emphasizes that the data employed to configure the chatbot will remain exclusive to the client company’s usage, with no application in the training of other models.
GPT-3.5 Turbo: A Remarkable Milestone
Introduced earlier in the year by OpenAI, GPT-3.5 Turbo stands as a family of AI models boasting notable capabilities. It boasts the capacity to process up to 4,000 tokens concurrently—a twofold improvement over its predecessors. Early testers have reported that by fine-tuning GPT-3.5, they managed to reduce their query lengths by a substantial 90%.
Affordable Customization and Usage
In terms of costing, the training of GPT-3.5 Turbo comes at $0.0080 per 1,000 tokens, while its usage incurs a fee of $0.0120 per 1,000 input and output tokens.
In a competitive landscape, Microsoft also presents GPT-based models that can be tailored via AI Builder and Power Virtual Agents, notes NIX Solutions. Additionally, OpenAI’s roadmap includes plans to offer configurability for GPT-4 in the upcoming fall season.