> US relied too much on secrets and menufactured inefficiency You are replying t... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		cscurmudgeon 6 months ago \| parent \| context \| favorite \| on: DeepSeek releases Janus Pro, a text-to-image gener... > US relied too much on secrets and menufactured inefficiency You are replying to a thread with the DeepSeek CEO saying the opposite (e.g., DeepSeek built upon transformers, Llama, PyTorch, etc.)

johnnyanmac 6 months ago [–]

I was talking about the US and the marketing of insane server racks of GPUs "required" to run popular LLM models. Did I misinterpret something?

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact