Artificial Intelligence thread

sndef888

Captain
Registered Member
So basically, a side project with a budget of 5 million destroyed western plans for technological supremacy

Only question I have (as someone who's not really into AI), is this possibly just a fluke or can we safely expect further market leading developments from Chinese AI teams?
 

gabriel.shenton

New Member
Registered Member
Been thinking.

With Deepseek coming out, doesn't that mean the whole AI thing hyped by techbros and H1B Indians is basically a huge ass scam?

Like why do we need all this money and immigration when some literal who guys back in China did it for just $5 million on some old GPUs?

Honestly Deepseek implications are not been discussed hard enough given this fact.
I believe it’s mainly used by the US to increase demand for the USD in the future , as the world slowly moves towards renewables, there is a structural less demand for USD, so now US is building out AI infrastructure, access of which tired to USD.

Also with regional payments infrastructure setting up around the world, lesser demand for USD.

Maybe a way to funnel the trillions printed in the last few years to the techbros
 

Fatty

Junior Member
Registered Member

Certain people more knowledgeable than me have also noted that some of the design choices that Deepseek have made are choices that would be negative if they were in fact GPU-rich like a certain Dylan is claiming.

also does anyone know where Deepseek is hosting their model to serve via their API? I hear they are using Huawei’s GPUs
 

OptimusLion

New Member
Registered Member
China Telecom releases "complex reasoning big model" TeleAI-t1-preview: can solve the "Nine Chapters of Arithmetic"

According to reports, TeleAI-t1-preview uses reinforcement learning training methods, and by introducing thinking paradigms such as exploration and reflection, it greatly improves the accuracy of the model in complex problems such as logical reasoning and mathematical deduction


001ZzMwgly1hxx5qp41eqj60u00g70x402.jpg

Please, Log in or Register to view URLs content!
 

tamsen_ikard

Junior Member
Registered Member
So I have a question:

How much do chips still matter in the AI race? And how much will the lack of top performing chips hinder China's future AI developments?

Neural network training is an "embarassingly parallel" (Yes that's a technical term) task. Which means you can break down the training task into tiny small fragments and run them on any hardware without any slowdown. So, it is possible to use millions of gaming GPU to train AI models instead of using H100s. Yes, it might consume more electricity and such but its still a cost effective solution. So, no, you don't need the most cutting edge chips. But you do need lots of chips.
 
Top