Calculating the task efficiency of Large Language Models and study on potential of eXtended Intelligence

抄録

With the rapid development and progress of the information society, the spread of artificial intelligence is also accelerating, and its types are quite diverse. Artificial intelligence can perform a variety of tasks in a short amount of time, and it dramatically increases human productivity. However, as artificial intelligence continues to develop, there are concerns that humans will eventually be replaced by artificial intelligence and jobs currently performed by humans may lead to people’s unemployment. Thus, amidst the discussion regarding the development and use of artificial intelligence, the concept of ‘eXtended intelligence (XI) ’ has emerged today. In this study, we focus on large language model and calculate how we can improve efficiency by using LLMs in clerical tasks. First, we calculate how we can improve efficiency when humans use LLM in clerical tasks such as summarizing or writing compared to when they do not use it. Based on the calculated LLM efficiency, we will conduct a preliminary study on the potential of LLM as a XI.