OpenAI Unveils GPT-4 Customization for Businesses, Enhancing Fine-Tuning on Company Data

Key Takeaways

• OpenAI has unveiled a new fine-tuning option for GPT-4o.
• The move expands customization options for OpenAI’s models.
• A growing number of businesses are embracing custom AI.

Fine-tuning in AI refers to an additional stage of model training intended to improve an AI agent’s performance in specific tasks. Developers offer fine-tuning options to enable customizable, domain-specific chatbots. However, the option wasn’t initially available for OpenAI’s most advanced Large Language Model (LLM), GPT-4o.

As of Tuesday, Aug. 20, however, OpenAI has listed  GPT-4o as one of the models it makes available for fine-tuning, equipping it with a degree of customizability that has been absent until now.

OpenAI Embraces Fine-Tuning

While the lightweight version of OpenAI’s flagship model, GPT-4o, was previously available for fine-tuning, the firm initially held off on opening up the full-sized version to customization. 

The firm has positioned the option as ideal for users who want to set a specific style, tone, or format for chatbot outputs, improve reliability within a specific domain, or teach the model new skills that are hard to articulate in a prompt.

Customers must upload their data to OpenAI’s servers to fine-tune a model. Depending on the size of the dataset, the process takes roughly an hour or two.

Although GPT-4o has multimodal capabilities, users can only fine-tune the model with text-based data.

Growing Market for Custom AI

With fine-tuning options available from major AI developers including OpenAI, Google and Anthropic, a growing range of corporate AI customers are embracing custom chatbots.

Use cases include customer service agents and sector or business-specific AI assistants requiring specialist domain knowledge. 

Of course, do-it-yourself fine-tuning isn’t the only option for AI customization. There is also a growing range of pre-specialized solutions catering to specific industries.

Platforms like Microsoft’s Copilot for Finance and Aveni’s dedicated model for financial services point to the huge opportunity presented by custom AI, which is seeing developers race to secure a slice of the burgeoning market for corporate AI services. 

Overcoming Security Concerns

While fine-tuned AI models create many new business opportunities, handing over proprietary data to companies like OpenAI for training raises important security and compliance concerns.

Of course, it’s difficult to envisage companies ever entrusting their most valuable commercial secrets to third parties. Many businesses have policies that prevent them from processing certain data through cloud services like chatbot APIs.

However, in an interview with CCN, Articul8 CEO Arun Subramaniyan observed that “we have gone past the stage of saying cloud is less secure than on-prem.” In fact, he said it is often more secure as cloud providers employ dedicated experts to protect their systems from threats.

He said a growing number of companies are starting to use automatic log analysis to take advantage of cloud-based generative AI solutions without sacrificing security compliance. Rather than running audits at defined intervals, companies can now automatically analyze cloud data flows, flagging any issues as they arise and enabling a much higher degree of security.

As businesses embrace custom AI, tools like log analysis can help increase trust and overcome some of the challenges inherent to the cloud-centric AI distribution model.

About the Author

James Morales

Although his background is in crypto and FinTech news, these days, James likes to roam across CCN’s editorial breadth, focusing mostly on digital technology. Having always been fascinated by the latest innovations, he uses his platform as a journalist to explore how new technologies work, why they matter and how they might shape our future.

See more

[email protected] LinkedIn Twitter