NVIDIA has recently introduced a new tool called NeMo Guardrails, which aims to help developers ensure that their generative AI applications are both appropriate and safe. This new tool provides companies with the ability to enforce three different types of limits on their proprietary large language models (LLMs). These include setting “topical guardrails” to prevent the AI from addressing subjects that it has not been trained on, safety limits to ensure that the AI pulls accurate information, and security limits to ensure that the AI connects only to safe applications.
NeMo Guardrails works with all types of LLMs, including the popular ChatGPT, and can be easily used by software developers without the need for them to have machine learning expertise. Moreover, since the software is open-source, it can be integrated with all the tools that enterprise developers currently use.
NVIDIA has incorporated NeMo Guardrails into its existing NeMo framework for building generative AI models, which business customers can gain access to through the company’s AI Enterprise software platform. NVIDIA has made the product open-source to contribute to the developer community’s efforts towards AI safety.
The introduction of NeMo Guardrails comes after some of the most high-profile generative AI models have come under scrutiny for their tendency to “hallucinate” information. This is where the AI generates false or incorrect information. Google’s chatbot, for example, made a factual error during its first public demo.
NVIDIA believes that the tool will help businesses keep their smart services aligned with safety, privacy, and security requirements, ensuring that these engines of innovation stay on track. The release of this new tool further demonstrates NVIDIA’s commitment to making AI more accessible and safer for businesses of all sizes.
If you want to learn more about how NeMo Guardrails works, NVIDIA has published a blog post that offers a deep dive into the software and provides information on how to get started with it.
