对抗低质内容（Slop）的战争
大家早上好！我希望你们今天都充满活力，因为我需要你们的支持。我有点紧张，但我们会一起度过。今天，我在这里要宣布对“低质内容”（Slop）宣战。
什么是 Slop？
我们先来讨论一下这个话题。每个 AI 工程师都有一个秘密。我已经和一些个人朋友分享过这个观点，我现在直接分享给大家。
- 第一届峰会：我们知道 AI 工程师将会是一个重要的角色。
- 第二届峰会：我们将范围扩展到领导层。
- 第三届峰会：我们意识到必须集中精力在模型实验室，这也是为什么今天世界顶级的实验室都聚集在这里。
我们通过 AIPMs 和 AI 设计师扩大了 AI 工程的影响范围。通过代码峰会，我们开始关注精选主题。如果今年最重要的主题是“编码”，但我今天不专门谈论编码。你们在接下来的时间里会听到很多关于编码的内容。
所以，请大家给我 5 分钟时间来谈谈“低质内容”。
我们做得很好，对吧？就像大家认为的那样，我认为“低质内容”与数量和质量有关。我思考的问题是，我们如何发展这个社区、这个行业和这个活动，同时保持高质量？
这也是我们非常关心的问题。我们在座位安排和演讲者选择上投入了很多精力。
- 我们正在与“低质内容”作斗争。这其实是牛津英语词典的内容，它甚至是 2024 年年度词汇的候选词，可惜输给了“脑腐”。
但我对牛津词典有一点小意见，因为他们说“低质内容”是通过人工智能生成的。
- 低质内容：低质量、缺乏真实性或不准确的内容。
- 但 AI 不一定带来低质内容：无论是人类还是 AI 都可能产生低质内容。
低质内容的例子
你应该自己也见过类似的情况。我想举几个例子。
- 互联网俚语：低质内容的反义词是“kino”。我从 Paul Rambles 那里学到了这个概念。
- 当我制作 Sora 视频时，我会和 Sam Holman 制作非常无聊的视频。
- 而其他有创造力且擅长工作的人制作 Sora 视频时，会出现猫玩弄挖掘机的场景。
同一个工作室可以制作出不同的内容。Netflix 制作了 K-pop 恶魔猎人，也有 Electric ST。
- 不同模型：Slop 可以由不同模型产生，不予评论。
Slop 可以退化为 SLO（低质内容），而 Kino 也可以退化为 SLO。如果你是趋势的早期参与者，当它是新鲜和新的时，这很好。
如果你认出另一张图片，说明你太过网络化了。
对抗低质内容的战争
[清嗓子] 好吧，关注的人不多。嗯，大家回去做作业吧。
显然，我会顺便嘲笑一下《权力的游戏》，因为这是一个类似的情况。就像低质内容无处不在一样。
它由人类和 AI 共同生成。你明白了吗？
[清嗓子]
好吧。同样的创业理念可以是优秀的，也可以是低质的。
当我在 AI 工程师峰会发表我的第一篇主题演讲时，我们实际上使用过一个 AI 幻灯片公司。
我加载了同一个幻灯片，但最近它关闭了，因为它实际上是一个低质的公司。
有关于“Vibe Coding”的不同看法，我认为其中一个比另一个要好得多。
我认为这些是我们必须应对的紧张关系之一。
我们的一位演讲者稍后也会提到这一点。
我认为 这两张图都是指数图表，但一个感觉更真实，一个更像是低质内容。我希望人们能研究一下为什么会出现这种情况。
总之，我们处于非对称战争中，而“低质内容”……
我认为 U 最接近的匹配法则是 Brandolini 定律，它实际上指出：
生产某物的能量需求比传播它所需的能量高一个数量级。
所以，我们需要创造一个合适的法律。
因为大家都知道，生成 token 的成本每年都在下降 100 到 1000 倍。
所以，这是 Fix 的反低质内容定律：
对抗 SLO（低质内容）所需的“品味”比生产它所需的品味高一个数量级。
外面有很多低质内容，我们需要提升世界上的内容，因为这就是人类所坚持的。
我认为有一个积极的信息：你可以用 AI 来对抗低质内容。
我很自豪能够运行一个名为 AI 新闻的副项目，它是唯一一个在没什么内容时告诉你不要阅读它的新闻通讯。
谢谢。哦，非常感谢。
你也可以提示 SLO（低质内容）。
下一位演讲者是 Mahash 和 Barry。
我发现他们在技能集中有一个提示，承认了低质内容并告诉云不要产生低质内容，结果确实显著改善。
代码低质内容
我们听说过代码如何创建技术债，两个工程师可以创造出相当于 50 个工程师的工作量。
或者说，在更严肃的情况下，你可以开始暴露数百万用户的私人数据。
所有我提到的内容都发生在今年。
我有点像用这个主题演讲来总结。
并且，就像开玩笑一样，即使有人说：
“哦，我的模型可以自主运行 30 到 60 小时。”
听起来有点低质，因为你也没有说：
“好吧，代码好不好？”
你只是在说它运行了多长时间。
所以，就像没有代表权就不要征税一样，你也不想在没有责任感的情况下享有自主权。
最近我一直在研究用 AI 对抗代码低质内容。
所以，这是来自一些人的引语：
半自动价值死亡，你可以在那里保持人类的注意力，并与机器融为一体，以便解决最难的问题。
而对于可以商品化的东西，你可以让它更异步。
你可以查看更多关于它的细节。
被低估的是我的其他关于代码地图的工作，我最近用 AI 完成了这项工作，以扩展基于代码的理解。
这也是对抗低质内容的一种方式。
你可以和楼下的认知团队交谈，他们可以向你展示更多细节。
最后一件事，我总是想提及的是计算机使用趋势。
我认为计算机使用在去年左右首次亮相，距今已经有一年了。
但它现在真的非常好。
它可以自主操作最复杂的应用程序，包括 ID。
所以，我认为这真的很令人兴奋，你们应该使用它来对抗低质内容。
我们用它来做网站，这里有一个例子，我们用 Devon 来自动化网站更新。
最后，我昨天从这次会议上学到的东西是：
你也可以使用子代理来对抗语境轮换。
而且，我认为这是我观察到的最大主题之一。
如果你们想从这次会议中获得一些收获……
我个人和 AIE 最大的亮点之一是与 Greg Brockman 聊天，他总是提倡模块化的概念。
你可以保持人类设计的清晰界限，让 AI 代码负责中间的一切。
所以，这些都是想法，但我只有一个信息，想压缩给大家：
“不再低质（No More Slop）”。
你的老板告诉你：
“我希望在季度末前有更多代码。”
你怎么回答？一起说：不再低质（No More Slop）。
你在打一场非对称战争。情况很糟糕。
你有一个测试不足的版本，可能会让你的公司感到尴尬。
对于真的很想推进它的人，你会说什么？
不再低质（No More Slop）。就是这样。
你的推特算法想要参与博取注意力的诱饵，并迫使你对广大公众撒谎。
你会说什么？
就是这样。
我希望你们有一个很棒的会议，让我们为**不再低质（No More Slop）**而努力。谢谢。热烈欢迎！
[音乐]
Original English
morning. How's everyone doing?
Good.
I'm going to need a lot of energy for
this talk, so please back me up. I'm
very nervous. Uh but we'll get through
this. I'm declaring war on slop today.
Let's talk about this. Every AIE has a
secret. I I've told this to uh some
folks that are personal friends and I'll
just show show the secret. Now the first
summit we had the secret which was we
knew that the AI engineer was going to
be a thing. Second summit we extended it
to leadership. Third summit we realized
that basically we always needed to
concentrate on model labs and that's why
you see um all all the all the top tier
labs here today um world's fair we
started expanding the TAM of of what AI
engineering uh is is affiliated with
with AIPMs and AI designers and with the
code summit as Jed just talked about we
really started focus on curation and
focusing in on a theme and if there's
one theme that's really matters this
year is it's coding but I'm not here to
talk about coding the rest of the day
you're going to hear about coding. So
just just indulge me 5 minutes to talk
about slop.
Um we we've done really well, right?
Like so so like I think like slop is
sort of associated with quantity and
quality. And I think like that's
something that I'm really trying to
think about as well like how do we grow
this community, grow this industry and
grow this event uh with the the same
kind of taste and high quality that
you've come to expect. And this is
something that, you know, I hope
hopefully you guys can see that we care
a lot about uh in curating all of you
coming here and all of the speakers that
you're about to see. Um we're in a war
against slop. Um this is actually uh the
Oxford English dictionary. Uh this is a
a candidate for the 2024 word of the
year. It lost to brain rot. [laughter]
But um but slop is slop is pretty good.
And I think it's it's probably gone even
more of an issue this year than last
year. Maybe it will win this year. I
have an issue with Oxford though because
they did us dirty by saying slop is is
generated using artificial intelligence.
The other part I I agree with. It's
lowquality, inauthentic or inaccurate.
But it doesn't take AI to be lowquality,
inauthentic or inaccurate.
Any human or AI can be an agent of slop,
right? You've seen this yourself. I'm
going to indulge me over a few examples.
By the way, I got this uh if you're not
familiar with like sort of internet
slang, the opposite of slop is kino. Um
and I got this idea from Paul Rambles
who uh like you know when I do Sora
videos, I do really boring Sora videos
with me and Sam Holman. When other
people who are actually creative and
good at their job do Sora videos, they
do cats playing digs.
Uh slot can be produced by the same
studio. There's K-pop demon hunters by
Netflix and there's electric st by
Netflix.
Slav can be produced in terms of
different models. No comment.
Slav can be SLO can something that's
keyno can degenerate into SLO, right? If
you're early on the trend and you're and
you you're starting that and it's and
it's fresh and new and that's great. U
if you're if you recognize the other
image, you're too online.
[clears throat] Okay, not enough people
recognize that image. Um go do your
homework. Um yeah, and obviously I'm
just going to throw in a dig at Game of
Thrones because this is the same same
thing, right? Like slop is everywhere.
It's generated by humans and AI. You get
it? [clears throat]
Okay. Um the same startup idea can be
keen versus slop. When I first presented
the my first keynote at AI engineer
summit, we actually used to uh which is
sort of like an AI slides company. Uh I
loaded I loaded up the same slide deck
and it was gone uh recently because uh
it was it was actually uh sort of uh
closed as a company. Um there's there's
you know different takes on vibe coding
and I think one of them uh is much
better than the other and I think like
this these are just like the tensions
that we have to navigate. um one of our
speakers later on as well meter. Um I
think it's really interesting that both
of them are exponential charts but one
of them feels more keeno and the other
is more slops and I think like I would
really like to have people investigate
why um so let let me let me just skip
through basically we're in an asymmetric
war and slop um I think u the closest
law that I found that matches this is
Brandolini's law which actually states
the amount of energy needed to refuel
is an order of magnitude bigger
than needed to produce it. Right. So we
need to co coin an appropriate law as
well. Um because you know like the the
the cost of generating tokens is is
dropping by 100 to a thousand times
every single year.
Um so this is I guess Fix's law of
anti-slav the amount of taste needed to
fight SLO is in order manually bigger
than that needed to produce it. Right?
There's so much low taste out there. We
need to elevate uh what's out there in
the world because that's what we stand
for as humans. Um, I think there's
there's a positive message. You can use
AI to fight slop. Um, I'm proud I'm
proud to run as a side project AI news,
which is the only newsletter that tells
you not to read it when there's nothing
going on. Thank [laughter] you. Oh,
appreciate that. Um, you can also prompt
SLO. The next speaker is um Mahash and
Barry. Um I found this in the in the
sort of prompt in the in the skill set
that they that they put they put where
they actually acknowledge slop and tell
cloud not to produce slop and it
actually improves significantly from
left to right. Um what about code slop
right we hear about code um creating
tech where you can sort of two engineers
can create a tech debt of 50 engineers
or you know on a more serious note you
can start exposing private data of
millions of users and this this all
happened this year. Everything I'm
mentioning all happened this year. I'm
kind of using this keynote as a as a way
of recapping. Um, and it, you know, just
to be spicy a little bit, even people
who are saying things like, "Oh, my
model can go up to 30 to 60 hours uh
autonomously." Well, it feels a bit
sloppy because you're also not saying,
"Well, was the code good or not?" You're
just saying how long it went. So, in the
same way that you have no taxation
without representation, you don't want
autonomy without accountability.
Um something I've been working on more
recently is that using AI to fight code
slop as well. So uh this is from the a
bunch of people quoted this yesterday of
the semi-ync value of death where you
can sort of keep human attention and
mind meld with the machine in order to
work on the hardest problems whereas the
stuff that's commoditized you can make
it more async. So you can check out more
on that details. What seems to be less
appreciated is my is the other work on
code maps uh which I which I more
recently done where we actually use AI
to scale codebased understanding which
is also a a way to fight slop and you
can uh talk to the cognition folks uh
downstairs uh um who who can who can
show you more in detail. Um the the last
thing I always want to shout out as well
is is this trend of uh computer use. I
think computer use kind of debuted it
this time last year uh with Enthropic.
But uh it's it's getting really really
good now guys. It can really
autonomously operate the most complex
apps including an ID. So I think that's
really uh exciting and you should
probably use that to fight slot. We use
it for the website and here's an example
of us using Devon to automate the the
sort of website up website updates. Um
and finally something I learned from
this conference yesterday is that you
can also use sub agents to fight context
rot. Um and I think that is one of the
biggest themes of of uh that I'm
observing as well. If you want to take
away something from this conference
um and I also one of the biggest
highlights of the year for us as AIE and
myself personally was chatting with Greg
Brockman who always uh preaches the
concept of modularity where you can sort
of keep clear boundaries on what is
human designed and let the AI code
everything in between. Um, so these are
all ideas, but I just have this one
message that I want to comp compress
down to you today that I want you to say
with me. No more slop. Yeah.
Your boss tells you, "I want more lines
of code in by the end of the quarter."
What do you say to that? Say it with me.
No more slop.
You're fighting an asymmetric war. This
is how bad it is, right? You have an
insufficiently tested release that that
is potentially embarrassing to your
company. What do you say to people who
really want to push it?
No more slop. Exactly. [laughter]
Uh your your Twitter algorithm wants
engagement bait uh and is sort of you
know forced and telling you to to to lie
basically to the to the broad public.
What do you say to that?
Exactly. That's it. Uh I hope you have a
great conference and let's let's hear it
for uh not having any more stuff. Thank
you. Heat. Heat. [applause]
[music]