David Sacks: Anthropic Wants to Kill Open Source AI in America

Watch on YouTube

Now Playing

Transcript

32 segments

0:00

I know people don't have a lot of

0:00

sympathy for Chinese companies. That's

0:02

fine. I'm not defending Chinese

0:03

companies. I'm defending American

0:05

developers who need to be able to use

0:07

everything in the public domain. Let me

0:08

give you an example. Cursor rolled out

0:10

its new product composer two. They were

0:12

able to post train that model using

0:14

Kimik 2.5 on their own proprietary

0:17

coding data. They started with a Chinese

0:20

open source model and then they used

0:21

their own data and they came up with a

0:23

new derivative product. This is the way

0:25

that open source works. You take things

0:27

that are in the public domain, you fork

0:29

them, you make them your own. And by the

0:31

way, once it's in the public domain,

0:33

it's not a Chinese model anymore. No

0:35

data is going back to China. Nothing is

0:37

going back to China. An American company

0:39

has taken open source contributions in

0:41

the public domain, made it their own,

0:42

and then developed their own model. And

0:44

if you say that American companies can't

0:46

do that or that somehow it's tainted

0:48

with IP theft, you are basically going

0:50

to put a dagger through the heart of the

0:52

entire American open source ecosystem.

0:54

And that is exactly what Anthropic wants

0:57

because they do not want to have the

0:58

competition.

Interactive Summary

Ask follow-up questions or revisit key timestamps.

The speaker argues that American developers must be free to use open-source models, including those originating from China, to foster innovation. They contend that once a model enters the public domain, it is no longer Chinese property, and restricting its use would severely damage the American open-source ecosystem while primarily benefiting incumbent competitors like Anthropic.