That's like blaming the collapse of the horse carriage industry on Carl Benz inventing the automobile.
Still, this is an epic bubble if it's popped. As someone else earlier in the thread correctly inferred by calling it weapons grade this is technological/economic nuclear first strike. The amount of devaluation this thing can cause is on similar level to an actual nuclear attack:
That was me, I was the one who called it weapons grade...
Here is the thing, by Deepseek publishing the methods and the weights, the cat is out of the bag, genie out of the bottle.... Even if US tries to TikTok Deepseek its too late
All the talk about limiting Open Source LLM by capping training compute is like horse already left the barn when now Huggingface is publically replicating R1
Even if US tries to regulate AI and ban self hosting or local inference, they cant force other nations to all do the same
Sure they can resort to GPU hostage holding with the AI Diffusion nonsense and putting countries on the blacklist tier but at the rate China EUV/GPU is catching up this will backfire long term for sure...
And so..... Their only one shot is a straight shot to ASI with a less than 2 year timeline but if R1 causes the bubble to pop prematurely then its game over.... Nvidia style
This might end up being the beginning of what ends up being the rug pull of the century
So not just a nuclear attack (heck even Japan survived two of those) but an existential civilization ending event. (if this ends up costing US the race to ASI)
Live by the sword, die by sword. US pinned all its hope on AI, went all in, and if DeepSeek manages to pop the bubble at the very top, well the resultant chain reaction is going to make the first Manhattan project look mild in comparison.
This is what total anniliation looks like