正文：
一项由普渡大学研究小组在5月初举行的计算机-人交互大会上提交的研究报告指出，ChatGPT在回答Stack Overflow上的编程问题时，错误率高达52%。这项研究检查了517个问题，发现ChatGPT提供的答案中有超过一半是不准确的。
尽管ChatGPT的错误率较高，但研究人员发现，在35%的情况下，人类程序员更喜欢ChatGPT的答案，认为其提供的信息更全面，且语言风格清晰。然而，在39%的情况下，人类程序员未能识别出ChatGPT提供的错误答案。
这项研究引发了对人工智能在编程领域准确性的关注，以及人类程序员对人工智能依赖程度的反思。随着AI技术的不断进步，如何平衡利用AI辅助与保持人类专业判断的准确性，成为了一个值得探讨的课题。
IT之家报道指出，这提醒了编程社区在使用AI工具时需要保持谨慎，并强调了对编程问题的回答需要经过严格的验证和审查，以确保最终解决方案的正确性。研究人员建议，在AI辅助编程的实践中，人类程序员应该对AI的输出进行复查，特别是在关键的编程决策上，以确保代码质量和程序的稳定性。
随着人工智能技术在各个领域的应用日益广泛，如何正确评估和利用AI的潜力，同时避免因依赖AI而产生的错误，将是未来技术发展中需要不断探索和解决的问题。
英语如下：
Title: AI Programming Assistants Often Fail: ChatGPT’s Error Rate Hits 52%, Yet 39% of Programmers Remain Unaware
Keywords: Errors, ChatGPT, Programmers
Content:
A study submitted by a research group from Purdue University at the Computer-Human Interaction conference in early May reveals that ChatGPT’s response to programming questions on Stack Overflow has an error rate of up to 52%. The research examined 517 questions and found that over half of the answers provided by ChatGPT were inaccurate.
Despite ChatGPT’s high error rate, the researchers found that in 35% of cases, human programmers preferred the AI’s answers, finding them more comprehensive and clearly written. However, in 39% of instances, human programmers failed to identify the incorrect answers provided by ChatGPT.
The study raises concerns about the accuracy of AI in the programming field and prompts reflection on the degree of dependence human programmers place on AI. As AI technology continues to evolve, the balance between leveraging AI assistance and maintaining the accuracy of human professional judgment becomes a topic worth exploring.
IT Home reports that this serves as a reminder to the programming community to use AI tools with caution and underscores the need for thorough verification and review of answers to programming questions to ensure the correctness of the final solution. The researchers suggest that, in the practice of AI-assisted programming, human programmers should double-check AI outputs, especially in critical programming decisions, to ensure the quality of code and the stability of the program.
As artificial intelligence technologies become increasingly prevalent across various fields, the correct evaluation and utilization of AI’s potential, while avoiding errors due to AI dependency, will be a continually evolving issue in future technological development.
【来源】https://www.ithome.com/0/770/553.htm
Views: 1