OpenAI is expanding its artificial intelligence offerings by allowing companies to fine-tune its ChatGPT-4 model in a move that could influence how businesses use AI.
The development represents more than just a technological upgrade; it indicates a move toward AI as a service, enabling companies to customize AI models according to their specific needs.
The introduction of fine-tuning capabilities for GPT-4o is a highly anticipated feature, especially for developers who have been eager to customize the model to suit their unique use cases.
OpenAI’s announcement includes an offer of one million free training tokens per day for its GPT-4o model for each organization until Sept. 23 and two million free training tokens per day until Sept. 23 for its GPT-4o mini, which is also available for fine-tuning.
Customizing AI for business needs
Fine-tuning allows businesses to optimize the GPT-4o model by training it on custom data sets. This customization can significantly enhance the model’s performance, enabling it to better understand and respond to complex, domain-specific instructions.
The ability to fine-tune AI’s tone, structure and functionality could have a big impact on its effectiveness in everything from coding and creative writing to customer service and technical support.
Related: Wyoming mayoral candidate wants to run capital city with AI bot
According to OpenAI, developers can achieve substantial improvements in model performance with as few as a few dozen examples in their training data sets. This means that even companies with limited data resources can fine-tune GPT-4o to meet their needs, making advanced AI technology more accessible than ever.
This fine-tuning update follows OpenAI’s signing of a multiyear paid agreement with Condé Nast, the media giant behind publications such as The New Yorker, Vogue and Wired.
News content partnerships
This partnership will allow OpenAI to train its AI models on Condé Nast’s library of content while also linking users directly to news stories from its outlets.
Although the financial specifics of the deal remain undisclosed, it marks a significant shift in how AI platforms can monetize and utilize media content.
However, this isn’t the first time the AI company has struck a deal to train its models on content from news media. In June, OpenAI and Time signed a multiyear content agreement to allow the former to train its models on more than a century of the latter’s work.
Magazine: AI may already use more power than Bitcoin — and it threatens Bitcoin mining