A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed a technique that allows large language models (LLMs) to generate more accurate and verifiable content by integrating information from external knowledge databases during the text generation process. This method, which contrasts with the standard approach of retrieving information only at the beginning of a query, interleaves retrieval steps throughout the generation of a long-form answer. Early testing suggests this ‘interleaved’ approach can significantly reduce the creation of factual inaccuracies, or hallucinations, a common challenge with current AI models. The research indicates this could lead to more reliable AI assistants for tasks requiring detailed, factual explanations, such as in technical support or educational contexts. For the full details, read the complete article at https://technologyreview.com/2024/07/18/1094851/mit-ai-retrieval-generation-hallucinations/