Positive users express hope for large-scale AGI benefiting humanity in 3-5 years from the DeepSeek chief's framing of AGI as industrial input, while negative users feel annoyed at repeatedly modifying logic for clients.
No Digg Deeper questions have been answered for this story yet.
I've been talking about this for a while Wenfeng's belief that AGI labs won't capture most value is not mere idealism but a coherent strategic position that's uniquely well-aligned with Chinese characteristics of the AI ecosystem. Intelligence is… just another industrial input.
Liang Wenfeng equating lowering collective cost of AI systems and mass diffusion to achieving AGI.
This will produce seethe in American labs that believe they're entitled to shares of the light cone What people miss is that Wenfeng isn't a random starry-eyed bro. He's not that young and he has a really good understanding of capitalism. He wrote a preface to Jim Simons' book. He made his career as a hedge fund CEO. He knows that electricity became a modest share of a vastly expanded economy; that funds earn small slivers of efficiently allocated capital. He's fine with becoming a "mere" provider of cognitive electricity. > What I often think about is whether something can make society operate more efficiently, and whether you can find a position you're good at in its industrial division chain. As long as the endgame is higher social efficiency, it's valid. Many things in between are just phases; focusing too much on them inevitably causes confusion. t. Liang Wenfeng, 2024.
People will read this as just "don't chase moats". I think it's more. Junzi Liang Wenfeng wants to have a negligible stake, but in an inconceivably large economy. This is the way of the Netsphere Engineer. Infrastructure for a civilization. 尔曹身与名俱灭， 不废江河万古流。
@teortaxesTex if intelligence becomes an industrial input, the prize shifts to whoever can turn cheap tokens into sustained demand for domestic compute
@goodhunt @teortaxesTex 希望能实现全人类的大规模AGI、但是还需要3-5年
@teortaxesTex He sees it for the signal transformation problem it is