NVIDIA NeMo Guardrails Open Source Software Helps Developers Add Security to AI Chatbots

NVIDIA NeMo Guardrails will help ensure that smart applications powered by large language models (LLMs) are accurate, appropriate, relevant and safe. The software includes all the code, samples and documentation a business needs to add security to the AI ​​applications companies use to generate text.

Today, when many industries adopt large language models (LLMs), generative AI is responsible for answering customer questions, summarizing lengthy documents, and even writing software and accelerating drug design. NeMo Guardrails aims to help users protect this new type of Security of artificial intelligence applications.

Powerful model, solid rails

The safety of generative artificial intelligence is the focus of the entire industry. NVIDIA designed NeMo Guardrails to be used with all large language models (LLMs), such as Open AI’s ChatGPT. The software allows developers to calibrate applications powered by large language models to be safe and maintain within a company’s domain of expertise.

NeMo Guardrails allows developers to establish three types of restrictions:

• Theme Safety FenceThis prevents the application from straying into undesired areas. For example, they prevent customer service assistants from answering questions about the weather.
• Security FenceMake sure the app responds with accurate and appropriate information. They can filter unnecessary language and enforce that only trusted sources be cited.
• security guardrailRestricts the app to only make connections with external connections that are known to be safe.

Even if you’re not a machine learning expert or a data scientist, nearly every software developer can use NeMo Guardrails to quickly create new rules with just a few lines of code.

Harness familiar tools

Because NeMo Guardrails is open source, it works with all the tools used by enterprise application developers. For example, it can run on LangChain, an open-source toolkit that developers are rapidly adopting to integrate third-party applications with the power of large-scale language models.

Additionally, NeMo Guardrails is designed to work extensively with applications driven by large language models, such as Zapier. Zapier, an automation platform used by more than 2.2 million businesses, directly demonstrates how users can integrate artificial intelligence into their work.

NVIDIAavailable as open source

NVIDIA is integrating NeMo Guardrails into the NVIDIA NeMo framework, which includes everything users need to train and tune language models using the company’s proprietary datasets.

The NeMo Framework is available as open source on GitHub. Enterprises can also access full enterprise support services through NVIDIA AI Enterprise.

NeMo is also provided as a service, part of the NVIDIA AI Foundations, a set of cloud services for enterprises to create and run custom generative artificial intelligence models based on their own datasets and domain knowledge.

Using NeMo, South Korea’s leading mobile network operator built a smart assistant that has conducted 8 million conversations with its customers. A team of researchers in Sweden used NeMo to create large-scale language models that can help automate text functions in the country’s hospitals, government and corporate offices.

source: Tech Blog