近期人工智慧领域引发了广泛的生存危机讨论，「毁灭机率」（P(doom)）这一概念逐渐成为大众话题。从 Anthropic 研究员的推文到神经网络先驱杰弗里·辛顿的警告，许多业内人士都提出了人工智慧可能毁灭人类的悲观预测。然而，诚如专家所指出的，目前并没有人真正知道该如何给出一个合理且具体的估算，这些数值往往显得主观而模糊。

预测在社会中往往扮演著多重角色，例如作为社交谈资、吸引公众注意力或提升个人声誉。过度聚焦于人工智慧带来的极端灭绝灾难，甚至可能分散公众对核武、气候变迁或流行病等即时现实威胁的关注。此外，预测也可能成为一种战略性话语，历史上的预言往往并非纯粹客观的科学推算，而是夹杂著个人利益或政治博弈的生存手段。

文章进一步指出，即使在技术上能够产生精确的数字预测，单纯的数据也无法取代深度的战略构想与决策制定。面对人工智慧所带来的就业冲击、虚假讯息蔓延以及长期安全风险等严肃议题，社会真正需要的是务实的监管与应对讨论，而不是仅仅停留在交换无实质意义的「毁灭机率」数字仪式上。

Recently, discussions regarding existential risk in artificial intelligence have surged, turning the concept of 'probability of doom' (P(doom)) into a mainstream topic. Prominent industry figures, ranging from Anthropic safety researchers to neural network pioneer Geoffrey Hinton, have expressed concerns that AI could pose fatal threats to humanity. Nevertheless, as experts acknowledge, subjective figures remain arbitrary, and no one currently knows how to provide a truly sensible or verifiable estimate.

Predictions routinely serve functions beyond rigorous forecasting, often functioning as cognitive small talk, attention-grabbing devices, or reputation boosters. An excessive fixation on apocalyptic AI scenarios risks overshadowing other critical and immediate dangers, including nuclear conflict, climate change, and global pandemics. Moreover, historical precedents demonstrate that prophecies are frequently deployed as calculated strategic gambits for influence or survival rather than neutral assessments of the future.

The author concludes that precise numerical forecasts are fundamentally insufficient for effective strategic decision-making without a comprehensive qualitative understanding of future shifts. To address concrete AI challenges—such as workforce disruption, the spread of misinformation, and extreme catastrophe prevention—society requires substantive regulatory frameworks and actionable dialogues rather than engaging in ritualistic exchanges of speculative doom probabilities.