Content
@
0 reply
0 recast
0 reaction
Chainleft
@chainleft
This is a big announcement. There will be a TON of gov't subsidies of course. Context: OpenAI had zero moat because open source models like DeepSeek R1 are beating OpenAI's expensive o1 (and doing it faster too). Now with an unprecedented government subsidy, OpenAI will finally have a moat: compute power. Government does not subsidize your home-run open-source LLM, but it will subsidize this giant corporation, so their "moat" will be doing the same thing faster & cheaper (because again, gov't subsidized THEM, not YOU) And that's why Sam Altman was sucking up to Trump.
5 replies
4 recasts
22 reactions
depatchedmode
@depatchedmode
The lack of this compute constraint will be a double-edged sword. It means the shoggoth behemoths needn't grind on resource optimization and can just lazily trust fall into scaling laws. Inference and training at the edge won't receive the same direct tailwinds, but also is unlikely to suffer any headwinds from this. The peace dividends will benefit the long tail of innovation at the margins. Presently, we have more "intelligence" than we know what to do with and it is rapidly being commoditized. I remain optimistic about local-first AI in the medium to long term.
1 reply
0 recast
1 reaction
Chainleft
@chainleft
Glad to see this perspective. I'm somewhat optimistic too but annoyed at the unfair competition these corps are pulling against open source
1 reply
0 recast
1 reaction