Generative AI relies on vast amounts of data to generate responses to user queries. For instance, OpenAI's GPT-3 was trained on the massive CommonCrawl data set comprising 570 gigabytes and 400 billion tokens. However, these data sets, although enormous, are static snapshots that cannot respond to queries about events unfolding in real-time. AI responses can also include hallucinations, where information appears plausible but is not factual. According to Vectara's Hallucination Leaderboard, even the most advanced LLMs, such as OpenAI's, exhibit hallucination rates ranging from 1.5 to 1.9 percent.
Therefore, using LLMs alone faces two significant challenges: the answers can be outdated, and the responses can be inaccurate. To address these potential issues, companies can leverage data streaming to inject new information into their data set and deploy retrieval-augmented generation (RAG) to encode business data in a format compatible with generative AI.
RAG generates a dataset that can be searched for relevant semantic matches to a user query. Those matches are then shared with the LLM for inclusion in the response. The vector data set can be updated or expanded over time, ensuring that relevant and timely data is available for incorporation into responses.
Click here to read the original article published by InfoWorld | IDG Communications, Inc.
Please give your feedback on this article or share a similar story for publishing by clicking here.
