Miscellaneous News

9dashline

Captain
Registered Member
This Deepseek thing is bigger than DeepSeek. US can try and ban DeepSeek but the model weights are already out in the wild under free MIT license plus Deepseek published the papers that allow anyone to use the recipe to duplicate /replicate and get close to SOTA with minimal effort... which huggingface and many others are already doing... so US would have to ban ALL open source, or move the goalpost to controlling GPU instead, which it already did with the AI Diffusion Export Controls in the last days of Biden.

Its about AI diffusion and the proliferation of local and open source /open weight AI winning against closed source gated (StarGate) AI where everyone in the world has to pay for StarLink and a $200/$2000 perptual license fee to US via External Revenue Tax....

As long as open source can keep bridging the gap (less than 6 months apart) from SOTA closed source, then US plan for its AI as the new hegemony wont be viable....

Until a few weeks ago I was wondering why China hasnt used its main rugpull yet, now I know why. It doesnt need to play that card yet

Its not if but when China catches up in EUV and GPU/AI chip production. As DeepSeek.proven already, Nvidia doesnt have moat because Cuda doesnt have moat and ASML is rapidly losing its own moat....

Getting to ASI first was Americas final hail mary to turn the tide, arrest its declining hegemony and wrestle away China's impeding global supremacy.... but it has to get to ASI fast enough and maintain a lead time gap wide enough....

but Now Im not even sure it will be the US that gets to ASI first, nevermind whether or not it maintains a sufficent lead while doing so....
 

AssassinsMace

Lieutenant General

The usual suspects are trying to undermine DeepSeek with things that don’t really matter. Who cares how much it really cost. The only thing matters third parties are saying it works as much or better than the best US closed models and China is giving it to the world for free. What are the usual suspects angry about? How much money they lose from the world not using their closed-source model that they wanted to charge. It doesn’t matter how much productivity DeepSeek spurs in endless economic growth and development everywhere it’s being used. They’re angry because no longer can they make money personally and exclusively from people using their closed proprietary model.

They accuse China of being a super surveillance state. China has an intrusive social credit system knowing the business and every single step of anyone traveling in the country. China gathers data from every single one of their devices they produce without permission and yet they think China doesn’t have as much data as the US has to make their AI just as good as theirs?
 

iewgnem

Junior Member
Registered Member
This Deepseek thing is bigger than DeepSeek. US can try and ban DeepSeek but the model weights are already out in the wild under free MIT license plus Deepseek published the papers that allow anyone to use the recipe to duplicate /replicate and get close to SOTA with minimal effort... which huggingface and many others are already doing... so US would have to ban ALL open source, or move the goalpost to controlling GPU instead, which it already did with the AI Diffusion Export Controls in the last days of Biden.

Its about AI diffusion and the proliferation of local and open source /open weight AI winning against closed source gated (StarGate) AI where everyone in the world has to pay for StarLink and a $200/$2000 perptual license fee to US via External Revenue Tax....

As long as open source can keep bridging the gap (less than 6 months apart) from SOTA closed source, then US plan for its AI as the new hegemony wont be viable....

Until a few weeks ago I was wondering why China hasnt used its main rugpull yet, now I know why. It doesnt need to play that card yet

Its not if but when China catches up in EUV and GPU/AI chip production. As DeepSeek.proven already, Nvidia doesnt have moat because Cuda doesnt have moat and ASML is rapidly losing its own moat....

Getting to ASI first was Americas final hail mary to turn the tide, arrest its declining hegemony and wrestle away China's impeding global supremacy.... but it has to get to ASI fast enough and maintain a lead time gap wide enough....

but Now Im not even sure it will be the US that gets to ASI first, nevermind whether or not it maintains a sufficent lead while doing so....
Even if ASI is possible anytime soon it would never have been the US. ASI is by definition self-creating, which means by definition it's a product of Deepseek type intelligence and not brute force. America's lack of intellgence and need to brute force definitionally already precluded them from ASI
 

9dashline

Captain
Registered Member
I am not lacking confidence in the home team (except in soft power). I just hate sharing with people that want to harm my tribe. I think it's insane. I always think about the global tragedy of Caucasians copying Chinese technology especially gun powder and then raping the entire planet. I never want that to repeat again. I already gave the example of an upgraded India with Chinese robots.

You look at Caucasians. How did they stay so strong for so long? They held onto their advantages. That's the smart thing to do.

You are looking at the positives of democratized AI. Imagine the open sourced Chinese tech leads to a gunpowder 2.0 situation, where a Chinese technology is once again weaponized by Caucasians to enslave everyone.

re: quote
There's a big difference between having a moat for self protection and prosperity vs plundering the world like Caucasians.

If we don't share nuclear technology then why share AGI, which is the ultimate weapon. Deepseek is not AGI, but that's their goal.
R1 is not AGI/ASI...

Deepseek releasing R1 as open weights precisely helps China's odds of getting to ASI first, before the Americans.

If this doesnt make sense to you, I suggest you use R1 to help you figure out why.
 

ougoah

Brigadier
Registered Member
R1 is not AGI/ASI...

Deepseek releasing R1 as open weights precisely helps China's odds of getting to ASI first, before the Americans.

If this doesnt make sense to you, I suggest you use R1 to help you figure out why.

How does R1 being released as open weights help China get to ASI first?

I haven't played with it but have seen and read a bit about it. Can you just give the general flow of logic behind this claim at least?
 

Chevalier

Captain
Registered Member
If the Anglo west wants Chinese scientists back, they should start by offering up the heads and hands of all China hawks immediately, starting with banning, pottinger, pompeo Navarro etc.
Then, compensation for all those affected by the pogroms to be paid in perpetuity just as Germany must pay Jews in perpetuity.
Then, they should offer gold and palaces and body slaves of young nubile Americans and make it a constitutional rule that anyone who so much as harms a Chinese person will be killed and the surrounding areas bombarded with carpet bombs, delivered by the PLAAF and paid for by the American taxpayer and billionaire class from their confiscated billions.

Any Chinese scientist who willingly goes to America is a traitor and should expect nothing less than lingchi; the Americans want to enslave Chinese, the only response is that given to the Xiongnu.
 
Top