文本记录English

AI broke the one thing we can't fix

6m 41s1,121 字数164 segmentsEnglish

完整文本记录

0:00

A few weeks ago, Nikita gets on his own

0:02

platform and says, "In less than 90

0:04

days, iMessage, phone calls, and Gmail

0:07

will be so flooded with spam and

0:09

automation that they will no longer be

0:11

usable, and we will have no way to stop

0:13

it." Now, this isn't some random schmuck

0:16

posting conspiracy theories on Reddit

0:18

about how Chipotle gives you diarrhea on

0:20

purpose. No, this is a man who gets

0:23

professionally destroyed by this

0:25

every single day and he's basically

0:27

telling the rest of us to lube up

0:29

because it's coming for everyone. 2

0:31

weeks later he purges 1.7 million bot

0:35

accounts off X and the next day they

0:38

just respawn. See, bots used to be

0:40

simple. Now they have like a PhD in

0:43

natural language processing and a

0:45

Bitcoin wallet. And so someone asked him

0:47

why it's about to get worse. and his

0:50

answer was the one word no one wants to

0:53

hear. Open claw. Now, if you're not

0:56

familiar with OpenClaw, firstly, I envy

0:58

you. It's currently the highest starred

1:00

software project in GitHub history. It

1:02

has more stars than React. Jensen Wong,

1:05

the CEO of Nvidia, who's almost

1:07

certainly never tried OpenClaw or even

1:09

knows what it does, just called it the

1:11

most important open- source project in

1:14

the history of humanity. What OpenClaw

1:16

does is it lets any with a laptop

1:20

build an AI agent that can send emails

1:23

and texts and browse the web and make

1:25

phone calls and impersonate you. Some

1:27

guy had his open claw negotiate car

1:29

prices with dealers over iMessage while

1:32

he was at brunch. Nikita's point is

1:34

basically what happens when every

1:37

scammer in Nigeria and every crypto bro

1:40

in Miami and every horny catfisher in

1:42

Wyoming gets one of these. See, because

1:44

before you needed to actually know

1:47

to be a spammer, you needed to know how

1:49

to operate servers and write scripts and

1:52

read code. Now you just need a laptop

1:54

and the moral compass of a pharma CEO.

1:57

And the problem with spam is not just

1:59

that it's a nuisance. The Chinese

2:01

government, surprise, surprise, they

2:03

figure this out before anyone. The

2:05

Chinese government floods ex search

2:07

results with porn every time there's

2:09

some political unrest. So when there's

2:11

protests in China, the search results

2:13

for Chinese keywords suddenly become

2:17

nothing but tits and ass. It's no longer

2:20

censorship or blocking the information.

2:23

Now you drown the information out with

2:25

content that is 10 times more engaging.

2:28

Some guy in Beijing opens up X to look

2:31

up protest footage and instead finds

2:33

himself jerking off for 45 minutes. The

2:35

CCP is looking through his webcam like,

2:37

"Nice, we got him." You can't overthrow

2:39

the government while you're jerking off.

2:42

It's the great come wall of China.

2:44

Basically, we all live behind that wall

2:46

now. And the reply spam on X, I mean, it

2:49

used to be like hot girls in bikinis

2:51

saying, "Check out my profile." Now it's

2:54

AI generated hot girls with AI generated

2:57

bikinis going, "Check out my AI

2:59

generated profile," which links to an AI

3:01

generated crypto scam. We used to get

3:03

scammed by real people, damn it. The

3:05

worst part about modern AI content is

3:08

that every single piece of AI generated

3:11

content was specifically engineered at

3:14

the neural network level to be more

3:16

engaging than anything a human being can

3:19

ever write. There's a process called RHF

3:22

where they train these models by showing

3:25

humans two responses and asking them

3:27

which one do you prefer and then they

3:29

just reinforce whichever one is more

3:31

compelling. Every AI model has been

3:34

mathematically optimized to hold your

3:36

attention better than any human written

3:39

piece of content ever could. Google

3:41

traffic to actual publishers dropped 38%

3:45

last year in the US alone. The economic

3:48

model of the internet is collapsing. We

3:51

had a deal. We paid humans to write

3:53

genuine things and in exchange they got

3:56

ad revenue. That deal is now dead

3:59

because why would an ad network pay a

4:01

human when a bot generates more clicks?

4:04

The algorithm doesn't care. The ad

4:06

networks don't care. Nobody in the chain

4:09

between creation and monetization has

4:12

any incentive whatsoever to care if it's

4:14

real. You could see this playing out

4:16

everywhere. Of YouTube's 100 fastest

4:18

growing channels last year, nearly 10

4:21

were fully AI generated. Slop was the

4:24

Miriam Webster word of the year for

4:26

2025. We went from selfie to slop in

4:30

about 10 years. That's basically the arc

4:33

of the internet in a single sentence.

4:34

And I mean, look, the internet was

4:36

always full of spam. Let's not pretend

4:38

it wasn't. But at least the spam used to

4:40

be stupid. You could see it coming from

4:42

a mile away. The grammar was You

4:45

had like three different fonts in a

4:47

single sentence. Nobody fell for that

4:49

Almost nobody. And now the spam is

4:52

smarter than you. It writes better than

4:54

you. The spam went out and got a college

4:57

degree and you're still on your parents'

4:58

phone plan. And Nikita, poor, poor

5:02

Nikita. I mean, he's still in there

5:03

fighting. He added a dislike button to

5:06

replies last week. That's basically like

5:09

the forefront of spam fighting today. A

5:11

freaking dislike button. And so, you're

5:13

probably wondering, okay, so what

5:15

happens next? What's the play? But this

5:18

isn't that kind of video. There's no

5:20

solution to this. This isn't a problem

5:22

we just fix. This is just what the

5:24

internet is now. And it's only going to

5:27

get weirder from here. Your inbox is

5:29

going to be so full of AI generated

5:31

emails that are so personalized and so

5:34

well written that the only way to know

5:36

if something is real is if you already

5:38

know the person who sent it. If some

5:40

stranger emails you a perfectly crafted

5:43

pitch, it's fake. If you get a LinkedIn

5:46

message that seems weirdly relevant to

5:48

like your exact career situation, fake.

5:50

You get a DM from a beautiful woman who

5:52

seems genuinely interested in your

5:54

personality, brother. Come on. The old

5:57

internet wasn't perfect, but at least

6:00

smart people could recognize the spam

6:02

from a mile away. In the new internet,

6:05

you won't be able to tell anymore

6:07

because the thing that's trying to fool

6:09

you was built from the ground up to be

6:12

better at fooling you than you are at

6:14

not being fooled. It's not a fair fight.

6:17

It's not even close. And in the

6:19

meantime, the only thing that still

6:21

means anything on this internet is the

6:23

people you already trust. It's not going

6:26

to be the individual posts or the

6:28

platforms or the algorithms. It's

6:31

people, specific humans who have earned

6:34

it over time. That's it. That's all

6:37

that's left.

解锁更多

免费注册以访问高级功能

互动查看器

观看带有同步字幕、可调节叠加层和完整播放控制的视频。

免费注册以解锁

AI 摘要

获取由 AI 立即生成的视频内容摘要、要点和结论。

免费注册以解锁

翻译

一键将字幕翻译成 100 多种语言。以任何格式下载。

免费注册以解锁

思维导图

将字幕可视化为交互式思维导图。一目了然地了解结构。

免费注册以解锁

与字幕聊天

提出关于视频内容的问题。直接从字幕中获取由 AI 驱动的答案。

免费注册以解锁

从您的字幕中获得更多

免费注册并解锁交互式查看器、AI 摘要、翻译、思维导图等。无需信用卡。

    AI broke the one thing we can't… - 完整文字记录 | YouTubeTranscript.dev