Nvidia's Universal 'Guardrails' Prevent AI Chatbots from Hallucinating False Information

Nvidia has released its open-source "NeMo Guardrails" software, which acts as a censorship bot for apps powered by large language models. The software is designed to help chatbots stay on topic and prevent them from spewing misinformation, offering toxic or outright racist responses, or creating malicious code. NeMo works on top of older language models like OpenAI’s GPT-3 and Google’s T5, and is supposed to work with "all major LLMs supported by LangChain, including OpenAI’s GPT-4." However, it remains unclear how effective an open-source guardrail might be in preventing AI from lying or cheating.