Nvidia introduces toolkit for safer and more secure AI text generation.
Nvidia has released an open source toolkit called NeMo Guardrails to make AI-powered apps that generate text and speech more accurate, appropriate, on topic, and secure. The toolkit is designed to work with most generative language models and can be used to prevent models from veering off topic, responding with inaccurate information or toxic language, and making connections to "unsafe" external sources. However, Nvidia acknowledges that the toolkit isn't perfect and won't catch everything.