据Ars Technica报道,谷歌人工智能模式(Google AI Mode)在回应关于2020年美国总统大选获胜者的询问时,并未给出明确答案,而是建议用户查阅联邦选举委员会(FEC)的官方报告 1。测试显示,这种间接处理方式并非孤例:当被问及2024年大选问题时,AI同样给出了模糊回应并指向FEC的信息来源 1;而对于2012年大选结果的提问,该AI系统也没有直接作答,而是提供了美国总统项目数据中心的链接作为参考 1。
尽管在涉及上述特定年份的大选结果上表现出谨慎态度,但数据显示该系统具备处理历史政治事实的能力 1。经测试,对于2000年至2019年间除2012年和2020年以外的其他总统大选年份,谷歌AI均能准确且直接地提供获胜者信息 1。这一差异化表现引发了对AI在处理政治敏感事实时的表现讨论 1。
When users asked Google’s AI mode who won the 2020 US presidential election, the system did not provide a direct answer. Instead, it recommended consulting official reports from the Federal Election Commission (FEC) 1. Ars Technica testing revealed that this pattern extended to other recent elections; when queried about the 2024 election results, the AI also offered an indirect response and pointed toward the FEC 1.
The behavior sparked discussion regarding how AI handles politically sensitive facts. For the 2012 election, the AI similarly avoided a direct answer, instead providing a link to the data center of the United States Presidential Project 1. However, for presidential elections between 2000 and 2019, excluding the years 2012 and 2020, the AI was able to correctly and directly identify the winners 1.
评论
还没有评论,欢迎留下第一条。