aichannel

DeepSeek R2 is launching soon, according to a Chinese tech media outlet.

Early reports describe a major drop in AI model costs and significant progress in hardware and training.

Source: a leaked article from JiuYanShe (韭研社) .

🧵 https://t.co/npTzp5RFbm

👋 Jan

https://x.com/jandotai/status/1916373755196281142?s=46&t=Il5NrMeCjY0zjgXBkcDWtA

Do you have any idea why China would open source their models?

What is the game theory around lowering their cost for inference? 

How does it benefit them when they're competing w the US?

china is ahead on manufacturing hardware at scale, so this is their way of commoditizing their complement

summoning the swarm @operator. mod /dacc, aixcryp.to

🙏This was the answer I was looking for

china is ahead on manufacturing hardware at scale, so this is their way of commoditizing their complement

https://gwern.net/complement

Couldn't that be said equally towards the US? 

We've seen multiple US models lower costs for inference + reasoning due to China's deepseek R1 model being open sourced

Weakening US dominance. China keeps the good stuff, kills US API profits, and breaks the global stack, dressed up as global good.