Researchers at OpenAI have improved the factual accuracy of language models by fine-tuning GPT-3 with a text-based web browser. This allows the model to access and incorporate information from the internet when answering open-ended questions, potentially reducing errors caused by outdated or incomplete training data. --- Why it matters: This matters because it addresses a common limitation of large language models: their reliance on static training data that may not reflect real-world knowledge. By enabling these models to browse the web, they can stay up-to-date and provide more accurate answers. Source: https://openai.com/index/webgpt