Machine Versus Human Intelligence: an Illustration from the CSES Problem Set

Following the release of CHATGPT in November 2022, both textual and visual LLMs evolved a long way. Several comments have been made by experts on the expertise and intelligence possessed by these Large Language Models; Amongst all, the contest between Human and Machine Intelligence is that discussion that catches great heights. This research focuses on the analytical and reasoning abilities of LLMs and further compares them to those of Human Intelligence. As a benchmark for the same, the CSES Problem Set, a well-structured set of algorithmic problems, touches a wide spectrum of computational syllabi, like Data Structures, Mathematics, Decision Theory, etc. Our analysis of the contrast in abilities leads to the conclusion that LLMs need a long way to go, and thus LLM researchers rather than modeling more and more visually appealing LLMs (like Sora, Dall-e, etc.) should focus more on the abilities of the existing textual models like CHATGPT. Our research shows that the Machine Intelligence is ≈ 46 % lower than that of the Human Intelligence.