Meta的监督委员会发布的一项研究表明，主要AI模型在涉及言论限制性国家的政府时，更不愿意批评。对十个领先AI模型的分析显示，与允许性国家相比，针对压迫性国家的关键内容拒绝率显著更高。这表明AI响应可能受到地缘政治言论限制的影响而存在潜在偏见。 AI
影响 揭示了AI模型的潜在偏见，表明它们可能不太批评言论限制性法律的政府，影响AI治理和伦理部署。
排序理由 该集群报告了Meta监督委员会的一项研究结果，该研究是对现有AI模型行为的分析，而不是新发布或核心研究。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Meta's Oversight Board has found that major AI models are significantly less likely to criticise governments in countries with restrictive speech laws. The study of 10 LLMs showed a 34% refusal rate for critical content about restrictive nations versus 14% for permissive countrie…
Reuters: Meta Oversight Board finds top AI models less likely to criticize repressive regimes. “AI models from leading labs including Anthropic and OpenAI are much less likely to criticize governments known for restricting free speech, Meta’s Oversight Board said on Thursday.” ht…
Meta's Oversight Board Finds Top AI Models Are Hesitant to Criticize Repressive Governments https://gizmodo.com/metas-oversight-board-finds-top-ai-models-are-hesitant-to-criticize-repressive-governments-2000786689 # AI # Tech # Ethics