Interesting post about the "Deepseek approach" vs the "OpenAI approach" w.r.t scaling: https://xcancel.com/_KarenHao/status/1883877986656825503
(It's from Jan 2025 though, so maybe already outdated).
I don't agree with everything in that thread, but I agree with the overall idea: using only more computing power for LLM training is a) bad for the planet and b) not even necessary.
#AI #LLM #Deepseek #OpenAI
ollibaba /
npub1u4…r6r4z
2025-05-29 09:45:40
Author Public Key
npub1u46xsczcv384rr96duhd4qvx4qzdlgfgh0nylh4e2ejt0vuzxwaqur6r4zSeen on
wss://relay.nostr.bandPublished at
2025-05-29 09:45:40Event JSON
{
"id": "4d94f8ddb4122cc1da2771a36b5428c6692bf02a8babe3fbfae6fc375f20acc5",
"pubkey": "e574686058644f518cba6f2eda8186a804dfa128bbe64fdeb95664b7b38233ba",
"created_at": 1748511940,
"kind": 1,
"tags": [
[
"t",
"ai"
],
[
"t",
"llm"
],
[
"t",
"deepseek"
],
[
"t",
"openai"
],
[
"proxy",
"https://chaos.social/users/ollibaba/statuses/114590478539961801",
"activitypub"
],
[
"client",
"Mostr",
"31990:6be38f8c63df7dbf84db7ec4a6e6fbbd8d19dca3b980efad18585c46f04b26f9:mostr",
"wss://relay.mostr.pub"
]
],
"content": "Interesting post about the \"Deepseek approach\" vs the \"OpenAI approach\" w.r.t scaling: https://xcancel.com/_KarenHao/status/1883877986656825503\n\n(It's from Jan 2025 though, so maybe already outdated).\n\nI don't agree with everything in that thread, but I agree with the overall idea: using only more computing power for LLM training is a) bad for the planet and b) not even necessary.\n\n#AI #LLM #Deepseek #OpenAI",
"sig": "a92d7b5aca6ce087a63925ba21eb954bf49d9c58316146ed43967b42f9c7c1b42ecfb990bb0282ed2ab7067745ec09da3d6a90857359a53e7ecfa3fdf8e5093b"
}