The author argues that the primary benefit of Large Language Models (LLMs) is their ability to allow computers and humans to interact in natural languages, rather than computer languages. LLMs can understand natural languages and translate them into a computer language suitable for the task at hand, making it easier for users to communicate with computers. The focus should be on improving the translation from natural to formal languages and giving LLMs access to existing software tools, rather than training them on vast amounts of data. The author emphasizes that traditional algorithms and databases are better suited for tasks like logic, calculations, and looking up facts, while LLMs excel in language and creativity. The author advocates for algorithms that provide accurate mathematical results and reliable databases for factual information, rather than relying on LLMs for these tasks. Ultimately, the breakthroughs of LLMs lie in enhancing language and creativity in computers, while facts and arithmetic have always been within the realm of traditional algorithms and databases.
Source link
Source link: https://medium.grid.is/let-large-language-models-handle-language-8d6830471f0e?source=rss——artificial_intelligence-5
Large Language Models mastering language; leave it to them. #NLP
