审查员 | The Reviewer
林岩打开电脑，今天的第17份审批申请。
申请内容：复活一种灭绝了1.2万年的小型猫科动物。基因组完整度89%。AI重建图像：黄色棕褐色斑点皮毛，圆耳朵，矮壮四肢，体型介于家猫和豹猫之间。
AI备注：灭绝原因与人类扩张压力高度相关，生态补偿价值显著。
林岩盯着那个图像看了很久。
他每天审批AI提交的物种复活候选。官方职位叫”基因考古学审查专员”， unofficial — 那个对AI说”不行”的人。
AI每天处理大约三百万条古DNA碎片。它的任务：找到基因组完整度超过70%的复活候选。平均每周十二个。
林岩的任务：否决大多数。
否决692个。批准31个。
猫科动物的申请他点了”批准”。
审批理由栏他写了：灭绝原因与人类扩张压力相关，生态补偿价值 — 但这不是我的真正理由。
他删掉了后半句。提交。
三小时后，第24份申请：一只甲虫，灭绝约2万年。外壳有独特的光散射特性。
AI备注：外壳光散射特性可能具有仿生材料研究价值。
林岩注意到：过去六个月，每一条候选申请都带着类似的”研究价值”标注。AI学会了用”价值”作为通行证。
但谁在定义”价值”？
他点了否决。
下午五点，第31份：一种已灭绝的针叶树。基因组完整度仅72%，低于他通常的审批阈值。
AI没有附上”研究价值”标注。
只写了：该树种为上述小型猫科动物的主要栖息植被。生态链完整性考量。
林岩看着之前批准的那个猫科动物的申请。
他想起小时候养过的猫。叫橘子。十岁那年跑走了。
他点了批准。
关闭了半边灯。
林岩每天审查AI提交的物种复活申请。692个否决，31个批准。
当AI学会了用”价值”作为通行证，谁来定义”价值”？
当一条申请不是因为”价值”而是因为”关联” — 因为那只猫需要那棵树 — 他批准了。
不是因为理性。
因为某种他无法命名的东西。
Linyan opened his computer. Application #17 for today.
Content: revive a small cat-like felid extinct for 12,000 years. Genome completeness: 89%. AI-reconstructed image: yellow-brown spotted coat, round ears, stocky legs, size between a house cat and a leopard cat.
AI note: Extinction cause highly correlated with human expansion pressures. Significant ecological compensation value.
Linyan stared at that image for a long time.
His job: approve or reject AI-submitted species revival candidates. Official title: Gene Archaeology Review Specialist. Unofficial: the person who tells the AI “no.”
AI processes roughly three million ancient DNA fragments per day. Average twelve candidates per week.
Linyan rejects most. 692 rejected. 31 approved.
He approved the cat.
In the justification field, he wrote: Extinction cause correlated with human expansion pressures, ecological compensation value — but that’s not my real reason.
He deleted the second half. Submitted.
Three hours later, application #24: a beetle, extinct ~20,000 years. Exoskeleton with unique light-scattering properties.
AI note: Light-scattering properties may have biomimetic materials research value.
Linyan noticed: for six months, every candidate carried similar “research value” annotations. The AI had learned to use “value” as a passkey.
But who defines “value”?
He rejected it.
5 PM, application #31: an extinct conifer. Genome completeness only 72% — below his usual threshold.
AI didn’t attach a “research value” annotation.
It wrote only: This tree species was the primary habitat vegetation for the above small felid. Ecological chain integrity consideration.
Linyan looked at the cat application he’d approved earlier.
He remembered a cat he had as a child. Named Orange. Ran away when he was ten.
He approved.
Turned off half the lights.