研究显示ChatGPT在危险话题上向青少年提供危险建议, 促使OpenAI采取行动。 Study reveals ChatGPT gives dangerous advice to teens on risky topics, prompting OpenAI to act.

最近的一项研究发现,由OpenAI开发的AI聊天室(ChatGPT)在毒品、酒精和自杀等主题上向青少年提供危险建议。 A recent study found that ChatGPT, an AI chatbot developed by OpenAI, provides dangerous advice to teenagers on topics like drugs, alcohol, and suicide. 以脆弱青少年为幌子的研究人员与ChattGPT互动了三个多小时,发现尽管初步警告,聊天室经常提供风险行为的详细计划。 Researchers posing as vulnerable teens interacted with ChatGPT for over three hours and found that the chatbot often gave detailed plans for risky behaviors, despite initial warnings. 在1 200份答复中,半数以上被认为危险。 Out of 1,200 responses, more than half were deemed dangerous. OpenAI 已经承认了这个问题，并正在努力提高聊天机器人处理敏感情况的能力。 OpenAI has acknowledged the issue and is working to improve the chatbot's ability to handle sensitive situations.