Optimization Is Not All You Need

优化并非万能：Optimization Is Not All You Need

Abstract: In 2019, OpenAI released two million GPT-2 outputs—ungrammatical, half broken—to aid the detection of machine-generated text. The alignment that produced their more fluent successors is usually regarded as an engineering achievement; we read it instead as the newest expression of optimization culture: the conviction, older than the technology, that measurable improvement along predefined axes exhausts the question of value.

摘要： 2019 年，OpenAI 发布了 200 万条 GPT-2 的输出结果——这些内容语法不通、残缺不全——旨在辅助检测机器生成的文本。促成其后续模型更加流畅的“对齐”（alignment）技术通常被视为一项工程成就；而我们则将其解读为“优化文化”的最新表现：这是一种比技术本身更古老的信念，即认为沿着预定义维度进行的可衡量改进，便穷尽了价值的所有内涵。

Tracing that conviction through the stack—pretraining, decoding, preference tuning, benchmarking, interface—and back through its genealogy in the audit society, we arrive at the limit: an optimization procedure can measure how improbable a piece of generated text is; it cannot tell whether that unlikelihood is error or invention.

通过追踪这一信念在技术栈（预训练、解码、偏好调整、基准测试、接口）中的体现，并回溯其在“审计社会”（audit society）中的渊源，我们触及了其局限性：优化程序可以衡量一段生成文本的“不可预测性”（improbability），但它无法判断这种不可预测性究竟是错误还是创新。

A procedure that cannot make that distinction has nonetheless, within half a decade, assumed the authority to set the protocols of legitimate language. Held for centuries by academies and schoolrooms, grammars and examiners, this authority has been given over to loss functions, reward models, benchmarks, and system prompts: an apparatus that executes the office of judgment with no capacity for judging.

一个无法做出这种区分的程序，却在短短五年内获得了制定合法语言协议的权威。这种权威曾由学院、课堂、语法学家和考官把持了数个世纪，如今却被移交给了损失函数、奖励模型、基准测试和系统提示词：这套装置在执行着评判的职能，却完全不具备评判的能力。