Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Study Finds That 52 Percent of ChatGPT Answers to Programming Questions Are Wrong

May 24, 2024 - futurism.com
A study by Purdue University researchers revealed that 52% of programming answers generated by OpenAI's ChatGPT are incorrect, raising concerns about the reliability of AI platforms. The study analyzed 517 questions from Stack Overflow and found that not only were many of the AI's answers incorrect, but they were also more verbose and inconsistent compared to human answers.

Despite these flaws, some programmers still prefer ChatGPT, with a small sample of 12 programmers preferring the AI's answers 35% of the time and failing to catch AI-generated mistakes 39% of the time. The researchers suggest that the AI's polite language and comprehensive responses may make its answers seem more convincing, leading users to overlook inaccuracies.

Key takeaways:

  • A team of researchers from Purdue University found that 52 percent of programming answers generated by OpenAI's ChatGPT are incorrect.
  • The study also found that 77 percent of the AI's answers are more verbose than human answers, and 78 percent suffer from inconsistency compared to human responses.
  • Despite these inaccuracies, many programmers seem to prefer ChatGPT's answers, possibly due to its polite language and comprehensive, textbook-style responses.
  • The study highlights that while AI platforms like ChatGPT are increasingly used, they still have significant flaws and can generate misinformation.
View Full Article

Comments (0)

Be the first to comment!