Complete News World

ChatGPT Succumbs to Programming Questions – M&K Advertising Week

ChatGPT Succumbs to Programming Questions – M&K Advertising Week

According to researchers at Purdue University, OpenAI’s ChatGPT AI chatbot is only suitable to a limited extent as an answer provider for programming questions. When creating code fragments for software projects, the tool gave a fairly large number of incorrect results in tests – especially with Stack Overflow.

Editorial Board

August 13, 2023


the Test series results Sober, because the chatbot didn’t answer even half of the questions correctly (48 percent). In addition, 78 percent of the responses were inconsistent compared to human responses at different scales – but they were more formal, more analytical, and the tonality was less negative. Almost every fifth result (22 percent) matches the answers given by humans.

However, ChatGPT brings an extra point in terms of comprehensibility, it is said. The bot is 65 percent understandable and the answers are easy to read. However, the abundance of irrelevant information provided remains a problem. (Pte/ Southwest)