USA just drew the red line in the AI war with China

[clutch]

American AI versus Chinese AI. The World may no longer have an option. You have to pick one can have it both ways. Who do you think the world should pick Chinese AI or American AI?

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

The United States is preparing to tell dozens of countries that they may have to choose between America and China in the rapidly escalating global race for artificial intelligence.

According to a draft State Department letter reviewed by Reuters, countries participating in Washington’s emerging AI coalition are being warned that they cannot simultaneously join competing Chinese initiatives. At the centre of the confrontation is Pax Silica, America’s attempt to build a trusted technology supply chain stretching from critical minerals and semiconductor manufacturing to data centres, advanced chips and artificial intelligence models.

But China is constructing an alternative. Beijing has launched the World Artificial Intelligence Cooperation Organization while Chinese open-weight models from companies including DeepSeek, Alibaba and Moonshot AI are spreading rapidly around the world.

In this GVS Deep Dive, Najma Minhas examines why Washington is moving from simply restricting China’s access to advanced technology towards building a global AI bloc, why Kazakhstan has suddenly become an important test case, how critical minerals and semiconductor supply chains fit into the competition, and why many countries may resist America’s demand to choose sides.

The biggest question may no longer simply be whether America or China builds the world’s most powerful AI.
It may be whose technological ecosystem the rest of the world chooses to join.


#AI #China #USA #ArtificialIntelligence #USChina #Geopolitics #Technology #PaxSilica #AIWar #Semiconductors #DeepSeek #OpenAI #Trump #XiJinping #GVSDeepDiv

 

[Ali_Baba]

while Chinese open-weight models from companies including DeepSeek, Alibaba and Moonshot AI are spreading rapidly around the world.

The open weight models make USA plans to "control" the development of Ai dead on arrival. The American model makes no sense, when you consider that soon you can locally host Ai models on your own servers, and build your own training data for your own uses cases using open Ai models.

While some of the frontier models are pushing the envelope, for most use cases the 2nd tier is more than good enough! At the rate that China is progressing, their frontier models will match the US models soon anyway.

 

[Hamartia Antidote]

The open weight models make USA plans to "control" the development of Ai dead on arrival. The American model makes no sense, when you consider that soon you can locally host Ai models on your own servers, and build your own training data for your own uses cases using open Ai models.

While some of the frontier models are pushing the envelope, for most use cases the 2nd tier is more than good enough! At the rate that China is progressing, their frontier models will match the US models soon anyway.

Corporate companies aren't going to let you randomly pick any LLM to generate things...they all have to be security vetted. Right now it is Codex and Copilot as pretty much the only viable ones.

While it is great saying this and that have high scores the bottom line is who is going to actually pay for it.

 

[SolarWarden]

I would rather want my AI overlords that will eventually enslave me and humanity to speak english instead of mandarin. So cheer for Murrcan AI to win the race to destroy humanity and not Temu AI.

 

[Ali_Baba]

Corporate companies aren't going to let you randomly pick any LLM to generate things...they all have to be security vetted. Right now it is Codex and Copilot as pretty much the only viable ones.

While it is great saying this and that have high scores the bottom line is who is going to actually pay for it.

View attachment 211151
View attachment 211152

That used to be also the case when companies insisted on paid for software on site, and now Linux rules the roost in the backend of every major company now and open source software sits on desktops as well.

Things will change, and Ai is new and people are trying to understand the technology and how to use it, so it is starting out as a restrictive model that will open over time.

As more open source models become available(they will), companies will gravitate towards those.

The Ai business model right now is the same as the dot.com business model.

 

[Hamartia Antidote]

That used to be also the case when companies insisted on paid for software on site, and now Linux rules the roost in the backend of every major company now and open source software sits on desktops as well.

Things will change, and Ai is new and people are trying to understand the technology and how to use it, so it is starting out as a restrictive model that will open over time.

As more open source models become available(they will), companies will gravitate towards those.

The Ai business model right now is the same as the dot.com business model.

It's simply too dangerous now. It used to be the days of worrying about putting a virus on your machine. Now you have software that can on the fly create a custom virus to attack your network...and this software is unscannable. There is no way to tell if it would do it or not unless you put it in a box and run thousands of tests on it...and even then you aren't sure.

If you put any LLM on a work machine or paste some code generated by a prompt window from a random website you can get fired. That's how serious it is. They put vetted LLMs you can use on a shared dedicated machine that can be disconnected.