中美AI差距只有六到八個月：全面封殺蒸餾也拖不了太多

核心論點 · 點時間戳可跳到原聲

原話 · 已逐字校驗

that plot, for example, just shows a 2X a month increase in Codex spending by researchers, and that does seem to me to be some evidence of they’re getting a lot of value out of this probably. I don’t think this is strong evidence that in six months we have a software intelligence explosion.

那張圖表明，研究員使用Codex的開支大約每月增長兩倍，這確實說明他們從中獲得了不少價值。但我不認為這是「六個月內出現軟件層面智能爆炸」的強證據。

JS Denain0:00

I think an analogy I have is, I think that junior PhD students will be 10X as productive, but from the advisor’s perspective, their research agenda will not proceed 10X as fast.

我的類比是：初級博士生的產出可能真的會快十倍，但從導師的角度看，整個研究議程未必會因此快十倍。

Nathan Lambert9:38

if you look at data centers in the US, there has been extremely fast build-out. If you had robots that were just literally able to substitute for blue-collar human workers, I feel like the financial incentives would be huge.

看看美國數據中心的建設速度就知道擴張有多快。如果機器人真的能直接替代藍領工人，我覺得財務上的激勵會大到足以推動建設。

JS Denain20:35

so currently if I’m saying six months gap, then in six months, it’s probably not literally 12 months, right? So, my guess is, yeah, you are at like eight or nine, something like that. You go from like six to nine maybe.

如果現在說差距是六個月，那六個月之後應該不會變成整整十二個月。我的猜測是大概會到八九個月左右，從六個月變成九個月左右。

JS Denain27:39

I think the Claude routers just give you a bunch of useful data to train on, and that will generally improve capabilities. I think that’s one explanation.

Claude的路由器能給你一大批有用的訓練數據，這本身就會普遍提升能力，我認為這是一種解釋。

JS Denain29:50

four-minute mile style effects of like, you show that something is possible at all and then people have the conviction to go down that route, don’t waste a ton of resources exploring a bunch of other things, is the thing you’re pointing at here.

這是一種「四分鐘一英里」式的效應：你先證明一件事完全可行，別人就有信心走這條路，不必再把大量資源浪費在探索其他方向上。

JS Denain38:27

there’s definitely clearly examples like this one, or really the Stolen Thoughts paper or something. It just seems like the safeguards weren’t good enough to prevent that from happening, even though there’s strong incentives for the companies to prevent it.

確實有像這次事件、或者Stolen Thoughts論文這樣明確的例子。看起來護欄並不足以阻止這些事發生，儘管公司有很強的動機去防止。

JS Denain48:46

數字與實體

術語

RSI遞歸自我改進

AI系統反過來加速AI研究本身的自我強化循環

ECIEpoch能力指數

Epoch AI用來衡量和比較模型能力水平的綜合指標

distillation蒸餾

用強模型的輸出數據訓練弱模型，從而快速獲得部分能力

four-minute mile effect四分鐘一英里效應

一旦有人證明某事可行，後來者不必再摸索就能跟進復現

jaggedness參差前沿

AI能力在不同任務上進步極不均勻，強弱交錯的狀態

Goodharted古德哈特化

一旦指標被當作優化目標，它就不再能反映真實情況

收聽指南