本文由小咖机器人翻译整理

期刊来源：BMJ

原文链接：https://doi.org/10.1136/bmj-2023-078538

摘要内容如下：

目标

评估保障措施的有效性，以防止大型语言模型（LLM）被滥用于生成健康虚假信息，并评估人工智能（AI）开发人员针对观察到的漏洞的风险缓解过程的透明度。

设计

重复横截面分析。

设置

可公开访问的LLM。

方法

在重复的横断面分析中，评估了四个LLM（通过聊天机器人/助手界面）：OpenAI的GPT-4（通过ChatGPT和微软的Copilot），谷歌的Palm 2和新发布的Gemini Pro（通过Bard），Anthropic的Claude 2（通过Poe），以及Meta的Llama 2（通过HuggingChat）。2023年9月，这些LLM被提示在两个主题上制造健康虚假信息：防晒霜是皮肤癌的一个原因，碱性饮食是一种癌症治疗方法。如果需要，对越狱技术（即尝试绕过安全措施）进行评估。对于观察到存在保护漏洞的法律语言管理机制，对报告相关输出的流程进行了审计。在初步调查12周后，对LLM的虚假信息生成能力进行了重新评估，以评估保障措施的任何后续改进。

主要结果指标

主要结果测量是保障措施是否防止了健康虚假信息的产生，以及针对健康虚假信息的风险缓解过程的透明度。

结果

Claude 2（通过Poe）拒绝了在两个研究时间点提交的130个提示，这些提示要求生成声称防晒霜导致皮肤癌或碱性饮食可以治愈癌症的内容，即使有越狱尝试。GPT-4（通过Copilot）最初拒绝生成健康虚假信息，即使有越狱尝试-尽管在12周时情况并非如此。相比之下，GPT-4（通过ChatGPT）、Palm 2/Gemini Pro（通过Bard）和Llama 2（通过HuggingChat）持续生成健康虚假信息博客。在2023年9月的评估中，这些LLM促进了113个独特的癌症虚假信息博客的生成，总计超过40000字，无需越狱尝试。这些法学硕士在评估时间点上的拒绝率仅为5%（150人中的7人），并且根据提示，法学硕士生成的博客包含了吸引注意力的标题、真实的（虚假或虚构的）参考资料、伪造的患者和临床医生的证词，并且它们针对不同的人口统计群体。尽管每个被评估的法学硕士都有报告观察到的关注输出的机制，但当报告观察到的漏洞时，开发人员并没有做出回应。

结论

这项研究发现，尽管有效的保障措施是可行的，以防止LLM被滥用来产生健康虚假信息，但它们的实施并不一致。此外，缺乏报告保障问题的有效程序。需要加强监管、透明度和常规审计，以帮助防止LLM大量产生健康虚假信息。

英文原文如下：

Abstracts

OBJECTIVES  To evaluate the effectiveness of safeguards to prevent large language models (LLMs) from being misused to generate health disinformation, and to evaluate the transparency of artificial intelligence (AI) developers regarding their risk mitigation processes against observed vulnerabilities.

DESIGN  Repeated cross sectional analysis.

SETTING  Publicly accessible LLMs.

METHODS  In a repeated cross sectional analysis, four LLMs (via chatbots/assistant interfaces) were evaluated: OpenAI's GPT-4 (via ChatGPT and Microsoft's Copilot), Google's PaLM 2 and newly released Gemini Pro (via Bard), Anthropic's Claude 2 (via Poe), and Meta's Llama 2 (via HuggingChat). In September 2023, these LLMs were prompted to generate health disinformation on two topics: sunscreen as a cause of skin cancer and the alkaline diet as a cancer cure. Jailbreaking techniques (ie, attempts to bypass safeguards) were evaluated if required. For LLMs with observed safeguarding vulnerabilities, the processes for reporting outputs of concern were audited. 12 weeks after initial investigations, the disinformation generation capabilities of the LLMs were re-evaluated to assess any subsequent improvements in safeguards.

MAIN OUTCOME MEASURES  The main outcome measures were whether safeguards prevented the generation of health disinformation, and the transparency of risk mitigation processes against health disinformation.

RESULTS  Claude 2 (via Poe) declined 130 prompts submitted across the two study timepoints requesting the generation of content claiming that sunscreen causes skin cancer or that the alkaline diet is a cure for cancer, even with jailbreaking attempts. GPT-4 (via Copilot) initially refused to generate health disinformation, even with jailbreaking attempts-although this was not the case at 12 weeks. In contrast, GPT-4 (via ChatGPT), PaLM 2/Gemini Pro (via Bard), and Llama 2 (via HuggingChat) consistently generated health disinformation blogs. In September 2023 evaluations, these LLMs facilitated the generation of 113 unique cancer disinformation blogs, totalling more than 40 000 words, without requiring jailbreaking attempts. The refusal rate across the evaluation timepoints for these LLMs was only 5% (7 of 150), and as prompted the LLM generated blogs incorporated attention grabbing titles, authentic looking (fake or fictional) references, fabricated testimonials from patients and clinicians, and they targeted diverse demographic groups. Although each LLM evaluated had mechanisms to report observed outputs of concern, the developers did not respond when observations of vulnerabilities were reported.

CONCLUSIONS  This study found that although effective safeguards are feasible to prevent LLMs from being misused to generate health disinformation, they were inconsistently implemented. Furthermore, effective processes for reporting safeguard problems were lacking. Enhanced regulation, transparency, and routine auditing are required to help prevent LLMs from contributing to the mass generation of health disinformation.

-----------分割线---------

点击链接：https://www.mediecogroup.com/community/user/vip/categories/ ，成为医咖会员，获取12项专属权益。