Thread - Nostr Hypermedia

Melvin Carvalho _@melvincarvalho.com 11 months ago

The game has changed. Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Model RL Revolution

Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Model RL Revolution

A Berkeley AI Research team led by PhD candidate Jiayi Pan has achieved what many thought impossible: reproducing DeepSeek R1-Zero's key technologi...

Replies (7)

npub16f67...wdy5 11 months ago

What a time to be alive!

StackSats.IO StackSatsIO@nostr.com.au 11 months ago

That’s nothing. I just trained an even better LLM on a bag of potato chips - come at me Techbros 😤

1 replies ↓

semisol semisol@nostr.land 11 months ago

I have heard you can rent LLMs for $30/hr with very high reasoning capabiltiies, self-checking correctness, extremely low hallucinations and even understanding emotions

1 replies ↓

StackSats.IO StackSatsIO@nostr.com.au 11 months ago

Yeah but you can’t eat the hardware when done

yaazer Yaazer@primal.net 11 months ago

Fair. But can it run doom?

1 replies ↓

StackSats.IO StackSatsIO@nostr.com.au 11 months ago

I can run doom on a single Dorito chip brother

Monkey Pox Denier 11 months ago

30$? That's not even the price of a PSU. Do they run that on a casio watch or what? With technicians earning 5c/h? Article does not mention the hardware they used. Can someone explain this to me?