06/26/2025, 12:02 AM UTC
人工智能的谎言：生成式AI搜索引擎的准确性危机AI Porkies
➀ 《哥伦比亚新闻评论》研究表明，ChatGPT、Perplexity和Gemini等生成式AI搜索引擎对超过60%的提问给出错误答案，其中Grok 3错误率高达94%；
➁ 付费版AI工具表现比免费版更差，倾向于提供「自信但错误」的答案；
➂ 这些工具逃避原始内容引用、伪造链接，切断新闻网站流量，对内容生态造成威胁。
➀ A Columbia Journalism Review study reveals generative AI search engines like ChatGPT, Perplexity, and Gemini provided incorrect answers to over 60% of queries, with Grok 3 failing 94% of tests;
➁ Premium AI tools performed worse than free versions, often delivering overly confident yet inaccurate responses;
➂ These tools bypassed source citations, fabricated links, and disrupted traffic to original content providers, challenging digital content ecosystems.
---
本文由大语言模型（LLM）生成，旨在为读者提供半导体新闻内容的知识扩展（Beta）。
