Mahesh Banavar - Böcker

Large Language Models (LLMs) have disruptively changed the world of AI for good and their adoption is near universal. However, how many know that they have a big limitation while processing large numerical quantitative business datasets usually found in ERPs as 1000s of tables. LLMs cannot process 100s of spreadsheets or tables at one time and when they try, they either fail to run or generate inaccurate predictions at best.The authors of this book propose LNMs or Large Numerical Models as a parallel universe to LLMs. LNMs are designed and built for numerical datasets and they offer some significant advantages over LLMs such as very accurate predictions, no hallucinations, improvement in business outcomes and ability to deliver in a "cold start" environment. LNMs are vertically curated and can run on a CPU as opposed to energy guzzling GPUs or water consuming cooling systems that LLMs need.This book introduces LNMs, it's underlying structure and SXI. SXI is to LNM as GPT is to LLMs, the underlying core science and technology. The authors also present specific applications of LNMs in healthcare, fintech, wireless, supplychain, marketing campaigns. Finally, the authors introduce their current research area of LLNMs. LLNM combines both LLM and LNM and has significant potential advantages over either LLM or LNMs.