Diogo Almeida
① 他说过的话
66 条
这在数学上是显而易见的,但在经验上并不成立。
指向原始笔记的链接
It’s mathematically obvious, but it doesn’t empirically hold.
—— Diogo Almeida · [07:15]
而且拒绝显然就是一个类型错误。
指向原始笔记的链接
And refusal is just obviously a type error.
—— Diogo Almeida · [12:02]
这个 AI 同事版的「无马马车」,而不是去挖掘 AI 的全部力量。
指向原始笔记的链接
this horseless carriage of AI co-worker instead of unearthing the full power of AI.
—— Diogo Almeida · [12:53]
对我来说,我认为智能会更像一个数据库,而不是一个同事。
指向原始笔记的链接
to me, I think intelligence will be more like a database than a co-worker.
—— Diogo Almeida · [15:59]
我愿意把我们的模型想象成,UDP 之于 LLM,TCP 之于我们的模型。
指向原始笔记的链接
I would like to think of our model like UDP as LLMs and TCP as our models.
—— Diogo Almeida · [23:56]
怎么能解决数学界的千禧年大奖难题,却连最基础的工作都无法自动化?
指向原始笔记的链接
How can we solve Millennium Prize problems in math? But still not automate even the most basics of works.
—— Diogo Almeida · [29:51]
就像是,创造重新回到了菜单上。
指向原始笔记的链接
It’s like creation is back on the menu.
—— Diogo Almeida · [31:30]
我们部署模型后不会更改它们。
指向原始笔记的链接
We will not change our models when we deploy them.
—— Diogo Almeida · [48:11]
对我来说,不自欺于我们的模型有多聪明,是其中最重要的事情之一。
指向原始笔记的链接
To me, not shitting ourselves about how smart our model is is one of the most important things there.
—— Diogo Almeida · [53:50]
我相信这一周发生的事情最大的救命稻草将会是人们的代码库。
指向原始笔记的链接
I believe that the biggest saving grace of what’s happening this week will be people’s codebase.
—— Diogo Almeida · [61:47]
会是人类,还是代码?我推断答案是有很多个 9 的概率是代码。
指向原始笔记的链接
Will it be humans or will it be code? And I figured it was many nines of code.
—— Diogo Almeida · [118:25]
其实有人让我做过电梯演讲,而我往往絮絮叨叨讲不好,但我意识到我最喜欢的 Jev 电梯演讲就是:到底他妈的自动化都在哪儿?
指向原始笔记的链接
So, I was actually asked for like an elevator pitch, which I tend to ramble on and I don’t do well, but like I realized my favorite elevator pitch for JEV is where the fuck is all the automation?
—— Diogo Almeida · [02:23]
我真正想要的是智能软件。与其自动化软件工程,我想拓展软件本身能做的事,让那些本该可自动化的东西真正变得可自动化。
指向原始笔记的链接
What I want instead is smart software. Like instead of automating software engineering, I want to expand what software itself can do such that things that should be automatable can then be automatable.
—— Diogo Almeida · [04:01]
编程本质上是把有价值的东西超精细地指定出来,然后无限复制它们。这太酷了,我只想把这件事做得更多。
指向原始笔记的链接
But like programming is like hyper specifying like valuable things and then infinitely replicating them. It’s so freaking cool and I want to just make that more.
—— Diogo Almeida · [04:25]
而且很离谱的是,AI 这么酷,而软件这 10 年却毫无变化。对我来说,这一点没人能把这两者调和起来。
指向原始笔记的链接
And also it’s wild that AI is so cool and software has been unchanged in 10 years. You know, like that to me, like no one can square this together.
—— Diogo Almeida · [06:15]
但目前,每美元智能是我的北极星指标。说清楚,这可能是错的。短期来看,每秒智能可能更有价值。
指向原始笔记的链接
But like intelligence per dollar is my North Star right now. And it could be wrong, just to be clear. Intelligence per second might be more valuable in the short term.
—— Diogo Almeida · [08:22]
我真的觉得那个模型有相当大的机会成为 AGI。而当它没成为的时候,那就是我整个世界观崩塌的时刻。
指向原始笔记的链接
I really thought that that model had like a decent chance of being AGI. And when it didn’t, that was like when my whole world came crashing down.
—— Diogo Almeida · [18:54]
我觉得 GPT-3 在那个时候其实是相当有校准的。但因为是人类在评估模型有多好,它看起来非常好,因为人类是裁判,但我们一直在优化那个裁判,而不是自动化那一部分。
指向原始笔记的链接
I think GPT-3 was actually quite calibrated back in that day. But because humans evaluate how good the models are, it looks really good because they are the judge, but we’ve been optimizing that judge instead of the automation part.
—— Diogo Almeida · [20:52]
就像,你真的要告诉我数学已经被解决了,或者说甚至两年前,GPQA 那种 Google 搜索都答不了的问答已经被解决了,但我们仍然搞不定一个得来速,对吧?
指向原始笔记的链接
Like, are you really telling me that math is solved or like even like two years ago, GPQA, that Google proof question answering is solved, but we still can’t handle a drive-through, right?
—— Diogo Almeida · [21:35]
但它讲的是那种懒惰,愿意花 10 个小时去把一个 5 分钟的任务做成瞬间完成,而且再也不用做第二遍。
指向原始笔记的链接
But it’s about the laziness to spend 10 hours to do the five-minute task instantly and to never have to do it again.
—— Diogo Almeida · [23:11]
可靠性就是这个东西的本质。如果你不理解这一点,就很难做出一个只在跑分上模仿的产品。
指向原始笔记的链接
Like, reliability is what this thing is. If you don’t understand that, it’ll be very hard to make, like, a copycat that’s benchmarked.
—— Diogo Almeida · [26:02]
我感觉可靠性的每一个九都会对所有人极其有价值,即使从市值角度看它不是最有价值的东西,因为它会直接启用新的应用。
指向原始笔记的链接
I feel like every nine of reliability is going to be so valuable for everyone, even if it’s not the most valuable thing market cap wise, because it will just enable new applications.
—— Diogo Almeida · [26:11]
它们不擅长语义。我会说它们在架构方面糟糕得难以置信。对我来说,架构是软件中最具人类创造性的部分。
指向原始笔记的链接
They’re bad at semantics. I would say incredibly bad at architecture. So like to me, architecture is like the most human creative part of software.
—— Diogo Almeida · [29:08]
但我认为 SaaS 将是整个 AI 游戏中最大的赢家之一。
指向原始笔记的链接
But I think that SAS will be one of the largest winners of like the whole AI game.
—— Diogo Almeida · [31:09]
想象一下,如果所有技术都做你所意。那不是科幻。看看 AI 有多聪明,对吧?
指向原始笔记的链接
Imagine if all technology just did what you mean. That’s not sci-fi. Look at how smart AI is, right?
—— Diogo Almeida · [41:52]
我用来介绍 Jev 的电梯演讲就是：那些该死的自动化到底都在哪儿。
指向原始笔记的链接
The elevator pitch I use for Jev is just where the f** is all the automation.*
—— Diogo Almeida · [01:47]
我觉得 AI 智慧得令人难以置信，但在你真正期望它有用的那些事情上，却又无用得令人难以置信。
指向原始笔记的链接
I think AI is just so unbelievably smart, yet so unbelievably useless at the kinds of things you’d really expect it to be useful for.
—— Diogo Almeida · [01:52]
至少 AI 实验室非常有动机不去诚实地回答这个问题。
指向原始笔记的链接
At least the AI labs are very incentivized to not answer that truthfully.
—— Diogo Almeida · [04:28]
所以我对这个答案及其细微之处的看法是：你为什么而优化，就得到什么——这对所有机器学习的事情都成立，说真的，对世界上所有的事情大概都成立。
指向原始笔记的链接
So my take on the answer and the nuance of it is that you get what you optimize for, which is just true of all things ML, kind of true of all things in the world, really.
—— Diogo Almeida · [04:55]
基本上所有强大的 LLM 都是为字符串优化的，而字符串是供人类或其他 LLM 消费的。
指向原始笔记的链接
Basically all the string LLMs have been optimized for strings and strings are meant to be consumed by humans or other LLMs.
—— Diogo Almeida · [05:18]
你知道，同样是 0% 的准确率，但你永远不会得到「把脸在键盘上乱按」这种情况，因为优化中的这种不对称性。
指向原始笔记的链接
You know, it’s the same 0% accuracy, but you never get the mashing your face on the keyboard because of the asymmetry in the optimization.
—— Diogo Almeida · [07:53]
而且我认为人们没有意识到 AI 从业者在设定他们期望的目标清单并植入模型方面有多大的主观能动性。
指向原始笔记的链接
And I don’t think people realize how much of an agency that AI practitioners have in setting their menu of desiderata into the model.
—— Diogo Almeida · [08:22]
但我相信，数据比算力重要得多。
指向原始笔记的链接
But I believe that data matters a lot more than compute.
—— Diogo Almeida · [10:02]
而比数据更重要的是，你需要正确的任务。
指向原始笔记的链接
And more important than data is you need the right task.
—— Diogo Almeida · [10:09]
所以我认为真正能创造一个新任务的人其实寥寥无几，这就是「你优化什么就得到什么」的那部分。
指向原始笔记的链接
So it’s actually a rare few that I think can actually make a new task, and that is the you get what you optimize for part of it.
—— Diogo Almeida · [10:36]
而在 RLHF 出现之前，互联网上确实根本不存在我们所要寻找的那种形态的人类反馈数据，因为我们不得不去创造它。
指向原始笔记的链接
And before RLHF existed, there was literally no human feedback data on the internet of the shape that we were looking for because we had to create it.
—— Diogo Almeida · [12:10]
我关于科幻的论点，尤其是和 LLM 相关的，我觉得这往往就是如今人们说 AI 时的意思，因为它似乎是智能的最大压缩，同时却有着最少的实用性。
指向原始笔记的链接
My argument with sci-fi is especially related to LLMs, which I think tends to be what people mean by AI these days, because it seems to be like the greatest compression of intelligence while having like the least utility.
—— Diogo Almeida · [16:08]
同时我也把自动驾驶看作一项巨大的工程胜利，而不一定是 AI 胜利。
指向原始笔记的链接
And also I see self-driving as a ginormous engineering win, not necessarily an AI win.
—— Diogo Almeida · [16:46]
我觉得这里愤世嫉俗的循环是：做一个 demo，融一笔种子轮或 A 轮，说你会把它做可靠，最后却从来没有把它做可靠，然后转而做一个 human-in-the-loop 版本的东西，而不是真正把这个任务自动化。
指向原始笔记的链接
My cynical loop here is make a demo, raise a seed or series A, say that you’re going to make it reliable, never end up making that reliable, pivot into a human-in-the-loop version of this thing instead of actually automating the task.
—— Diogo Almeida · [17:33]
那大约是六年前的事了，而客服仍然没有被解决。
指向原始笔记的链接
That is now roughly six years ago, and customer service is still not solved.
—— Diogo Almeida · [19:13]
人们不断尝试自动化得来速却不断失败，对此没有什么好的解释，除非得来速比未解的数学难题还难。
指向原始笔记的链接
People keep trying and failing to automate drive-throughs and there’s no good answer for that other than maybe drive-throughs are harder than unsolved math.
—— Diogo Almeida · [19:41]
而且总体来说，获得更高可靠性的方法是放大聚焦。
指向原始笔记的链接
And in general, the way to get higher reliability is to zoom in.
—— Diogo Almeida · [21:26]
但北极星一直是：如果 AI 真正、真正地为「对软件有用」而优化，它会长什么样？
指向原始笔记的链接
But the North Star has always been, what does AI look like if it’s really, really optimized for being useful for software?
—— Diogo Almeida · [23:09]
它们就是「有用性」的形态本身。
指向原始笔记的链接
They are literally the shape of usefulness.
—— Diogo Almeida · [25:15]
我其实觉得，如果有人告诉我们，我们做的是「针对任何东西的零样本通用分类器」，那会是 Jev 得到过的最高赞美。
指向原始笔记的链接
I actually think if someone told us that we were a zero-shot general classifier for anything, that would be the greatest compliment I’ve ever been given for Jev.
—— Diogo Almeida · [25:37]
而我相信他们在思想上相当正确，但他们一直被拖累，因为模型是为字符串优化的，而不是为他们想要的东西优化的，也就是程序化的使用。
指向原始笔记的链接
And I believe that they’re intellectually quite correct, but they’ve been hobbled by the models being optimized for strings instead of them being optimized for the thing that they want, which is programmatic use.
—— Diogo Almeida · [27:23]
而我的团队一直让我别再跟人讲这个，因为这会让我们更有可能遇到竞争。
指向原始笔记的链接
And my team has been telling me to stop telling this to people because it makes it more likely we have competition.
—— Diogo Almeida · [28:19]
我觉得人们就是没意识到他们是在为智能付费。
指向原始笔记的链接
I think people just don’t get that they’re paying for intelligence.
—— Diogo Almeida · [28:53]
我担心的是，如果人们玩的是这个模型的糟糕版本，他们可能会被整个子类别坑到，然后说：嘿，这类东西就是垃圾。
指向原始笔记的链接
I worry that if people play with bad versions of the model, they might actually get burned by the whole subgenre and say like, hey, like, like this kind of stuff is crap.
—— Diogo Almeida · [31:03]
你知道吗，我想成为的标杆是，让 AI 可靠到无聊透顶，就像 SQL 那样。
指向原始笔记的链接
You know, like, I want to be a paragon of making AI so reliable that it’s boring, like SQL.
—— Diogo Almeida · [32:26]
而且，我再怎么强调都不为过，可靠性才是人们付费买的东西。
指向原始笔记的链接
And also, I cannot emphasize enough, reliability is what people are paying for.
—— Diogo Almeida · [32:58]
说到这个，如果你想听一个关于我的有趣又尴尬的事，就是当时我有了这个想法、那个 aha 时刻击中我的时候，我在 OpenAI 写了一份文档，题目是「我们本来去年就能拥有 AGI」。
指向原始笔记的链接
On that note, if you want a fun, embarrassing thing about me, is at the time I had this idea and the aha hit me, I wrote a document in OpenAI entitled, titled, maybe not entitled, titled, we could have had AGI last year.
—— Diogo Almeida · [33:38]
如果我们用我们这点预算能做出比全世界其他人都更好的预训练，那全世界其他人都彻底完蛋了。
指向原始笔记的链接
If we could pre-train better than the rest of the world with the budget we’ve had, then the rest of the world is totally cooked.
—— Diogo Almeida · [43:39]
你知道，指数级增长的资源换来有争议的线性收益，看起来是非常次线性的。
指向原始笔记的链接
You know, exponentially more resources for debatably linear gains seems very sublinear.
—— Diogo Almeida · [43:57]
发表 RLHF 对世界来说可能是件好事。
指向原始笔记的链接
Publishing RLHF was probably really good for the world.
—— Diogo Almeida · [48:01]
任何 function calling API 都不应该存在，而不带一个针对函数本身的逻辑偏置。
指向原始笔记的链接
No function calling API should exist without a logic bias for the function itself.
—— Diogo Almeida · [49:35]
所以对我来说，RLCD 就是把正确的接口暴露给构建者，让他们可以获得想要的属性，而不只是对着 system message 祈祷。
指向原始笔记的链接
So to me RLCD is about exposing the right interface to builders so that they can get the properties they want without just praying to the system message.
—— Diogo Almeida · [50:16]
一个 ML API 的产品本身有很多门道，大多数人没有意识到，因为他们无脑地追随 MAU、月活用户之类的东西的梯度。
指向原始笔记的链接
There’s a lot to the product of an ML API that most people don’t realize because they mindlessly follow the gradient of things like MAUs, monthly active users.
—— Diogo Almeida · [52:13]
实际上，RLHF 和 RLVR 极大地破坏了模型的校准，因为要想把字符串输出好，你实际上需要极度过度自信。
指向原始笔记的链接
Actually, RLHF and RLVR destroy the calibration of the models immensely because in order to output strings well, you actually need extreme overconfidence.
—— Diogo Almeida · [56:42]
它并不像其他模型那样处于收益递减的极端平台期。
指向原始笔记的链接
And it is not in the extreme plateau of diminishing returns like other models are.
—— Diogo Almeida · [69:10]
我的猜测是，当他们最终把它放进分发中时，就发生了这种情况，因为你优化什么就会得到什么。
指向原始笔记的链接
My guess is this is what happened when they finally put it in distribution because you get what you optimize for.
—— Diogo Almeida · [70:48]
如果你的系统消息里有你整套业务算法，而逻辑却运行在别人的服务器上，这是很奇怪的。
指向原始笔记的链接
It’s weird for the logic to be running on someone else’s server if your system message has your whole business algorithm on it.
—— Diogo Almeida · [74:03]
而我愿意这样想：总有一天智能会像数据库一样，当你在代码中需要智能时就直接调用它。
指向原始笔记的链接
And I like to think of it as like one day intelligence will be like databases where you just call it when you need it, when you need intelligence within your code.
—— Diogo Almeida · [74:25]
KVCache 有点像到处使用全局变量，但如果你能对自己的状态聪明一些呢？
指向原始笔记的链接
The KVCache is a little bit similar to having global variables everywhere, but what if you could be smart about your state?
—— Diogo Almeida · [80:48]
慢慢地加入越来越多的逻辑，因为要在像 Claude Codes 和 Codexes 这样的东西里添加额外的 harness 逻辑非常困难，因为它们的核心循环就是一个 while 循环。
指向原始笔记的链接
That slowly add more and more logic because it’s very hard to add additional harness logic into like the Claude codes and codexes because their core loop is just a while loop.
—— Diogo Almeida · [82:37]
而且待在轨道上是一个特性，你知道的，你不会想去一个没有轨道的游乐园，或者至少没有轨道、安全带和所有那些东西你就玩不了最好玩的项目，对吧？
指向原始笔记的链接
And it is a feature to be on the rails, you know, like you don’t want to go an amusement park that doesn’t have its rails, or at least you can’t do the most fun stuff without like the rails and the seatbelts and all of that, right?
—— Diogo Almeida · [86:24]
② 出现在这些集
3 集
- 《Jev 背后的人:Diogo 谈为什么 AI 应该像数据库而不是同事》 — 作为嘉宾
- 《AI 这么聪明,自动化去哪儿了?Jev 把智能装进软件本身》 — 作为联合主持
- 《AI 为什么这么聪明，却干不了保险核保？》 — 作为联合主持
③ 他谈到的
点进去有真内容 —— 本页主要出口
Jev · TypeSafe · RLHF · OpenAI · RLCD · 可靠性 · RLVR · 分类器 · 校准 · ChatGPT