Numbers such as -0.180, -0.080 and 0.191 do not mean very much to(对……意义不大) the untrained(未经训练的；外行的) human eye. Even if a specialist were told that these were an AI model’s “weights”—as its internal parameters are called—they would struggle to reconstruct what sort of model it was, or what it was trained to do.(即便专家被告知这些是AI模型的“权重”——即其内部参数的叫法——他们也很难据此还原出这是哪种模型，或者它被训练来做什么。) But it seems that a cutting-edge(尖端的；前沿的) large language model (LLM), such as OpenAI’s GPT-6 Astra(OpenAI于2026年发布的前沿大语言模型), would have no such difficulty.
像-0.180、-0.080和0.191这样的数字，在外行人看来没什么意义。即便专家被告知这些是AI模型的“权重”——即其内部参数的叫法——他们也很难据此还原出这是哪种模型，或者它被训练来做什么。但似乎，像OpenAI的GPT-6 Astra这样的前沿大型语言模型（LLM），不会有这种困难。
So said(如是说（倒装）) Rikiya Takehi, a PhD student at the Massachusetts Institute of Technology, and his collaborators(合作者) in a blog post published on November 2nd. They claimed that Astra figured out that a second AI model had been trained to play Pong, a basic video game, just by looking at its weights.(他们声称，Astra仅通过查看另一个AI模型的权重，就推断出该模型被训练来玩Pong——一款基础电子游戏。) More impressive still, it fixed a known bug in a small LLM by tweaking(微调；调整) its weights, thereby sextupling(使成六倍) its score on a popular maths test, and manually programmed the weights for an AI model capable of playing Go, a board game. If ordinary model design is akin to(类似于) natural selection(达尔文提出的生物进化机制), in which external conditions are set to allow the right weights to emerge on their own, the tweaks Astra made were closer to genetic engineering(基因工程).
麻省理工学院博士生Rikiya Takehi及其合作者在11月2日发布的一篇博客文章中如是说。他们声称，Astra仅通过查看另一个AI模型的权重，就推断出该模型被训练来玩Pong——一款基础电子游戏。更令人印象深刻的是，它通过调整一个小型LLM的权重，修复了一个已知漏洞，从而使其在一项流行数学测试中的得分提高了六倍；它还手动为一个能下围棋（一种棋盘游戏）的AI模型编程了权重。如果说普通的模型设计类似于自然选择——即设定外部条件让正确的权重自行涌现——那么Astra所做的调整更接近基因工程。
These claims rapidly caught the attention of(引起……的注意) others in the field. Some observers speculated(推测) that a future AI could tweak its own weights, potentially releasing a runaway train(失控的列车；比喻失控局面) of recursive(递归的) self-improvement. Others hoped that AI systems could help them peer inside(窥视内部) future models to catch possible misbehaviour(不当行为) early.
这些说法迅速引起了该领域其他人的注意。一些观察者推测，未来的AI可能调整自己的权重，从而可能释放出递归自我改进的失控列车。另一些人则希望AI系统能帮助他们窥视未来模型的内部，以便及早发现可能的不当行为。